RESEARCH REPORT
Twisting space: are rigid and non-rigid mental transformationsseparate spatial skills?
Kinnari Atit • Thomas F. Shipley • Basil Tikoff
Received: 28 November 2012 / Accepted: 5 February 2013 / Published online: 20 February 2013
� Marta Olivetti Belardinelli and Springer-Verlag Berlin Heidelberg 2013
Abstract Cognitive science has primarily studied the
mental simulation of spatial transformations with tests that
focus on rigid transformations (e.g., mental rotation).
However, the events of our world are not limited to rigid
body movements. Objects can undergo complex non-rigid
discontinuous and continuous changes, such as bending and
breaking. We developed a new task to assess mental visu-
alization of non-rigid transformations. The Non-rigid
Bending test required participants to visualize a continuous
non-rigid transformation applied to an array of objects by
asking simple spatial questions about the position of two
forms on a bent transparent sheet of plastic. Participants
were to judge the relative position of the forms when the
sheet was unbent. To study the cognitive skills needed to
visualize rigid and non-rigid events, we employed four tests
of mental transformations—the Non-rigid Bending test (a
test of continuous non-rigid mental transformation), the
Paper Folding test and the Mental Brittle Transformation
test (two tests of non-rigid mental transformation with local
rigid transformations), and the Vandenberg and Kuse
(Percept Motor Skills 47:599–604, 1978) Mental Rotation
test (a test of rigid mental transformation). Performance on
the Mental Brittle Transformation test and the Paper Fold-
ing test independently predicted performance on the Non-
rigid Bending test and performance on the Mental Rotation
test; however, mental rotation performance was not a
unique predictor of mental bending performance. Results
are consistent with separable skills for rigid and non-rigid
mental simulation and illustrate the value of an ecological
approach to the analysis of the structure of spatial thinking.
Keywords Mental transformations � Rigid
transformations � Non-rigid transformations �Mental visualizations
Introduction
Spatial thinking ranges from concrete action planning to
abstract visualization of multi-dimensional space. While it
is possible that humans’ ability to visualize passing a plate
at dinner may to a first approximation be related to a sci-
entist’s ability to visualize the movements of continental
plates, it seems unlikely that a single cognitive process
subserves all concrete and abstract reasoning about the
range of spatial problems that confront a mobile symbolic
thinker. Spatial problem solving involves visualizing and
manipulating a broad class of spatial information—the
relations between locations, configurations, shapes, and
objects and how they can change over time (Newcombe
and Shipley in press). The ability to mentally represent and
transform spatial relations makes up one’s ‘‘spatial ability’’
(Carroll 1993) or ‘‘spatial skills.’’ Researchers in cognitive
This article is part of the special issue on ‘‘Spatial Learning and
Reasoning Processes’’, guest-edited by Thomas F. Shipley, Dedre
Gentner and Nora S. Newcombe. Handling editor of this manuscript:
Dedre Gentner.
K. Atit (&) � T. F. Shipley
Department of Psychology, Temple University,
1701 North 13th Street, Philadelphia, PA 19122, USA
e-mail: [email protected]
B. Tikoff
Department of Geology, University of Wisconsin-Madison,
Madison, WI, USA
B. Tikoff
Department of Geoscience, University of Wisconsin-Madison,
1215 W Dayton Street, Madison, WI 53706, USA
123
Cogn Process (2013) 14:163–173
DOI 10.1007/s10339-013-0550-8
science have endeavored to characterize the skills that
comprise spatial thinking and define the categories of
spatial skills; although there is general agreement about the
importance of spatial thinking and that there is not just one
skill (e.g., Guildford and Lacey 1947; McGee 1979;
Thurstone and Thurstone 1941), there is surprisingly little
agreement about the details of what makes up this skill set
(Caplan et al. 1985).
One suggested distinction in spatial skills that has been
agreed upon by many researchers is between two classes of
spatial transformations: (1) object-based transformations
and (2) egocentric perspective transformations—as
required during navigation (e.g., Hegarty and Waller 2004;
Kozhevnikov et al. 2006). This paper focuses on spatial
skills associated with object-based transformations that
could be used to reason about events—how objects change
(Shipley and Zacks 2008). Research on what defines spatial
ability regarding this specific class of transformations
remains unsettled because work that has attempted to
define spatial ability has focused on a relatively restricted
range of spatial information. Research has relied heavily on
a few measures (e.g., mental rotation), which sample only a
few of the many types of events human beings think about
(Gibson 1986; Shepard and Cooper 1986). This may in part
reflect the difficulty in constructing such measures, but it
may also reflect limitations in recognizing the breadth of
cognitive spatial skills that should be assessed. Using
extant measures to define categorical boundaries will only
work if the measures sample broadly from the spatial
thinking space. One notable program of work by Pani and
colleagues has sought to extend the territory of spatial
transformations by characterizing the processes involved in
sequences of mental rotation (e.g., Pani 1993; Pani and
Dupree 1994; Pani et al. 1995), as well as interactions
between rigid objects, such as reasoning about the pro-
jections of a shape onto a planer surface (Pani et al. 1996),
the orientation of the line defined by the intersection of two
surfaces (Pani et al. 1998), and the shapes resulting from
slicing the brain (neuroanatomy) (Chariker et al. 2011). As
there is no formal account of the universe of potential
spatial interactions among rigid objects, we are not sure
how extensively this impressive research program has
explored spatial thinking. One way to approach the prob-
lem of ensuring that research spans the breadth of spatial
thinking is to develop an account of the dimensions of
spatial cognition and ensure research covers this space.
Alternatively, one may analyze the types of spatial prob-
lems faced by humans and ask, does existing research
account for all (or at least many) of these problems?
A number of attempts have been made to characterize
the dimensions of spatial thinking by identifying com-
monalities in the cognitive mechanisms (e.g., Maccoby and
Jacklin 1974), identifying clusters in skill using meta-
analyses and factor analyses (e.g., Ekstrom et al. 1976;
Linn and Petersen 1985), using computational modeling
(e.g., Smith et al. 1982), and using neuroscience methods to
identify candidate biological substrates associated with
different types of spatial thinking (e.g., Mehta and New-
combe 1991; Milivojevic et al. 2003; Vogel et al. 2003).
Recently, Chatterjee (2008) proposed two broad dichoto-
mies of spatial thinking based on a review of psychologi-
cal, linguistic, and neuroscientific data, suggesting that
humans process intrinsic (objects’ shapes and part-based
representations) and extrinsic (relations among objects and
between objects and frames of reference) spatial repre-
sentations differently, and static (stationary) and dynamic
(moving) spatial representations differently. This low-
dimensional division of spatial thinking appears to
exhaustively categorize all spatial skills into the four broad
categories that involve the four types of spatial relations:
intrinsic static, intrinsic dynamic, extrinsic static, and
extrinsic dynamic.
In contrast to the breadth of the existing categorizations
of spatial skills, when we consider the types of events that
have been used to study spatial cognition, we find a rela-
tively narrow focus. Generally, the events employed to
study mental simulation of change by cognitive scientists
are rigid (transformations where distances between every
pair of points on an object is preserved) (e.g., Hegarty and
Waller 2004; Pani et al. 2005; Shepard and Metzler 1971).
Yet, humans can visualize events that involve complex
non-rigid changes. For example, we can readily imagine
the hood of a car before and after a collision. As a result of
the collision, the hood underwent a non-rigid transforma-
tion, where the distances among points on the hood chan-
ged as the metal crumpled.
In cognitive science, there is recognition that humans
can perceive non-rigid transformations (Gibson et al. 1978;
Gibson and Spelke 1983; Gibson and Walker 1984; Spelke
1982) and that the brain processes complex transformations
(e.g., Shepard 2001; Shepard and Cooper 1986; Shepard
and Metzler 1971), but there is almost no research on how
humans visualize complex non-rigid transformations, like a
car crash. To be sure, there are some notable exceptions to
this observation: Pittenger and Shaw (1975) studied rec-
ognition of shear and strain in aging, Hegarty et al. (2003)
studied reasoning about complex mechanical events
involving pulleys, and recently, Resnick and Shipley (in
press) studied expert geologists’ mental simulation of
brittle deformations (e.g., breaking) that involve multiple
translations of fragments. Aside from this handful of cases,
what we know about the mental simulation of non-rigid
events comes primarily from research on mental folding.
An extensive analysis of the research on mental rotation
and mental folding, which involves a sequence of piece-
wise rigid changes, concludes that there is likely partial
164 Cogn Process (2013) 14:163–173
123
overlap in the skills—the two skills share some cognitive
processes, but each also requires some unique mental
processes (Harris et al. in press).
From the perspective of wanting to study spatial cog-
nition as it applies to natural events, folding does not seem
representative of all the different types of transformations
available in the environment and only captures one aspect
of non-rigid changes, the fact that the change is not uni-
form within an object. No set of experiments has attempted
to study the simulation of more naturalistic non-rigid
events with continuously varying change. Here, we focus
on the simple event of bending. To study mental reasoning
about this event, we developed the Non-rigid Bending test,
which included items that required a non-uniform contin-
uous non-rigid transformation. We employ the test to try to
understand the relationship between mental folding and
mental rotation by studying reasoning about rigid and non-
rigid events. If reasoning about bending relates to measures
of reasoning about folding or breaking and not to measures
of rigid transformations (e.g., mental rotation), we may
conclude that rigid mental transformations are indeed
independent of those involved in reasoning about non-rigid
transformations in general. Alternatively, the skills needed
to reason about a sequence of piece-wise changes, such as
folding and breaking, might be distinct from those needed
to reason about continuous changes applied over an entire
array, such as bending or rotating. To further understand
how humans reason about non-rigid events, we looked to
see if skill in reasoning about one type of event predicted
skill on others. We tested participants reasoning about:
bending, breaking, folding, and rotating.
Methods
Participants
One hundred and seventeen college undergraduate psy-
chology students were recruited through Temple Univer-
sity’s Undergraduate Psychology Research Pool.
Participants provided informed consent and received credit
toward a research participation requirement for their
involvement in the study. Data from one participant were
discarded, because he did not complete the study. Data
from 116 participants were used in the analysis (38 men, 78
women, Mage = 20.60 years, age range 18–53 years).
Materials
Non-rigid Bending test
The Non-rigid Bending test was designed as an objective
test of a participant’s skill in mentally simulating bending.
For the test, 120 19.3 by 19.3 cm transparent plastic sheets
were printed with a black circle and star, shown in the two
examples provided in Fig. 1. The star was positioned on
the top half of the sheet, and the circle was positioned
10 cm from the bottom. A red line defined the bottom edge
of the sheet. The surface of the plastic sheet was defined in
one of two ways, either with random array of polygons or
with a 6 by 6 grid of lines. Half of the plastic sheets (60)
had a textured background (Texture Condition), with
polygons (1 by 2 mm—created using the ‘‘confetti’’ pattern
in Adobe Illustrator CS3) ‘‘sprayed’’ on the sheet with a
density of 10 polygons per square cm. The other half of the
sheets (Gridlines Condition) had 6 vertical and 6 horizontal
lines drawn 2.7 cm apart.1 An example of both types of
sheets is shown in Fig. 1. Each sheet was bent in three
ways to vary task difficulty (Appelle 1972; Rock 1973):
simple bend (the top was bent down to the level of the
bottom with no offset), moderate oblique bend (achieved
by bending a top corner down to the level of the bottom of
the sheet and offset by 10 cm from a simple bend), and
complex oblique bend (achieved by bending a top corner
down to the level of the bottom of the sheet and offset by
20 cm from a simple bend). Here, we refer to sheets bent
orthogonally as easy, sheets that had a moderate oblique
bend as medium, and sheets that had a complex oblique
bend as hard.
Three bends for each image yielded 180 stimuli for each
of the Texture and Gridlines conditions. Of these, 25
‘‘easy’’, 30 ‘‘medium’’, and 30 ‘‘hard’’ were selected at
random. For examples of the stimuli used in the Non-rigid
Bending test, see Fig. 2.
Participants were presented with the 85 stimuli twice
over the course of the task for a total of 170 trials. Trials
were presented in two blocks with a minute rest in between.
The stimuli were presented in a random order. For each
trial, the participants’ task was to determine what the sheet
would look like if it were unbent. Specifically, they were to
indicate whether the star would be to the left or right of the
circle if the sheet were flat (un-bent). Participants were
given 6 s to respond to each trial. The presentation of
stimuli and collection of data, accuracy and reaction
time (RT), were controlled by E-Prime (version 2.0).
1 Two separate conditions were designed to pick up potential
strategies that involved using a frame of reference within the bent
sheet to make the left–right judgments. For example, rather than
mentally unfolding the plastic sheet, a participant might try to
mentally construct an imaginary line parallel to the sides of the sheet
that extended from one form and then judge whether the other form
was to the left or right of this line. Such a strategy would have been
suggested by finding that the grid stimuli were easier than the texture
stimuli. No subjects spontaneously reported using this strategy, while
a number mentioned mentally unfolding the plastic sheet, and as
reported below, there was no evidence of a difference between the
two stimuli types.
Cogn Process (2013) 14:163–173 165
123
Participants were instructed to press the ‘‘p’’ if the star
would be to the right of the circle, and the ‘‘q’’ if the star
would be to the left of the circle. Each trial began with a
fixation cross presented for 1.5 s.
Paper Folding test
The Paper Folding test was administered using the proce-
dure described in Ekstrom et al. (1976) Kit of Cognitive
Reference Tests. In this test, participants were presented
with figures representing a piece of paper being folded one
to three times and then having a hole punched in it. The
task was to visualize what the paper would look like when
unfolded and to select the correct answer from one of five
answer choices. Following Ekstrom et al.’s (1976)
instructions, a participant’s score on the task was corrected
for guessing by subtracting one-fifth of the number of
incorrect answers from the number of correct answers.
Scores could range from -4 to 20 points. The test has two
parts with 10 questions each. Participants were given 3 min
to complete each half.
Mental Brittle Transformation test
The Mental Brittle Transformation test is a test of skill in
reasoning about breaking developed by Resnick and
Shipley (in press). A shortened version of the original task
was used in this study. The stimuli were created by
inserting a repeating character between each letter of a
familiar word, and then, the whole item was cut into pieces
which were moved in one of three ways: faulted, where the
cuts and displacement of parts were consistent with the slip
motion of a geological fault; random, where the fragments
were displaced in random directions; and exploded, where
the fragments were displaced radially away from a central
point. Resnick and Shipley (in press) found that using the
three types of changes offered a good range of difficulty in
identifying the original word. Participants were given
1 min to visualize what each word looked like before it was
broken. Participants were given the opportunity to view
each stimulus twice. Three faulted words, three exploded
words, and three randomly displaced words were pre-
sented. The items were presented in a random order. To
score the test, one point was given for each correctly
identified word. Scores could range from 0 to 9.
Mental Rotation test
Participants completed the Peters et al. (1995) paper and
pencil version of the Vandenberg and Kuse (1978) Mental
Rotation Test. For this test, the participant was presented
with five line drawings of 3D forms similar to those used
by Shepard and Metzler (1971). The target form is on the
left and four answer choices on the right. For each of the 24
problems, participants were asked to identify the two
choices that were identical but rotated versions of the target
form. Using Vandenberg and Kuse’s (1978) suggested
method of scoring, participants only received credit for an
item if both figures were identified correctly. Scores could
range from 0 to 24 points. The test has two parts with 12
problems each. Participants had 3 min to complete each
half.
Procedure
This study had a between-subject design. Participants were
alternately assigned to one of the two background condi-
tions of the non-rigid bending task: Gridlines (n = 57) and
Texture (n = 59). One person was tested at a time in a
closed and quiet room. After arriving for the study, the
A B
Fig. 1 Illustrations of the plastic sheets used to generate the stimuli in the Non-rigid Bending test: a textured background condition, b gridlines
background condition
166 Cogn Process (2013) 14:163–173
123
participant first completed the consent process. After giv-
ing consent, he or she completed the Non-rigid Bending
test, and then the three other tests. The Mental Brittle
Transformation test, the Mental Rotation test, and the
Paper Folding test were given in a random order. Partici-
pants averaged 50 min to complete the study (range
45–60 min).
Results
Initial analysis of the Non-rigid Bending test
To investigate the effect of the amount of bending on
performance on the Non-rigid Bending test, a one-way
ANOVA was conducted on all stimuli.2 Levene’s test for
homogeneity of variances was found to be violated for the
present analysis, F(2, 345) = 18.29, p \ .001, and there-
fore, a Kruskal–Wallis H Test was used instead. The
analysis revealed a significant effect of level of bending,
H(2) = 119.85, p \ .001. Bonferroni-corrected Wilcoxon
Signed Rank post hoc comparisons revealed that as the
level of bending increased in complexity, accuracy
decreased. Participants’ accuracy on the Non-rigid Bending
test was significantly lower on the ‘‘medium’’ trials
(M = .85, SD = .13) than on the ‘‘easy’’ trials (M = .92,
SD = .13), p \ .01, and lower on the ‘‘hard’’ trials
(M = .72, SD = .18) than the ‘‘medium’’ trials (M = .85,
SD = .13), p \ .001. This suggests that the Non-rigid
Bending test is a valid measure of non-rigid mental
transformation.
Reaction time on the Non-rigid Bending test
To investigate the effect of the level of bending on RT for
the Non-rigid Bending test, a second one-way ANOVA
was conducted. The analysis again revealed a significant
effect of level of bending, F(2, 345) = 17.76, p \ .001.
Tukey’s post hoc comparisons revealed that as the level of
bending increased, RT increased. Participants’ RT on the
Non-rigid Bending test was significantly greater on the
‘‘medium’’ level trials (M = 1.66 s, SD = .60 s) than on
the ‘‘easy’’ level trials (M = 1.38 s, SD = .57 s), p \ .01,
and greater on the ‘‘hard’’ trials (M = 1.85 s, SD = .66 s)
Fig. 2 Examples of stimuli from the Gridlines background condition
in the Non-rigid Bending test: a is an example of an ‘‘easy’’ trial—an
image of a sheet with an orthogonal bend, b is an example of a
‘‘medium’’ trial—an image of a sheet with a moderate oblique bend,
c is an example of a ‘‘hard’’ trial—an image of a sheet with a complex
oblique bend
2 To investigate the relationship between the background pattern and
the amount of bending on performance on the Non-rigid Bending test,
a mixed-design ANOVA with a within-subject factor of level of
bending (easy, medium, hard) and a between-subject factor of
condition (Gridlines, Texture) was conducted. Mauchly’s test indi-
cated that the assumption of sphericity had been violated
(v2(2) = 56.37, p \ .001), and therefore, degrees of freedom were
corrected using Greenhouse-Geisser estimates of sphericity (e = .72).
The analysis revealed no effect of condition, n.s. An analysis of RT
revealed a similar pattern: There was no effect of condition and no
significant interaction, n.s. As there was no effect of background
pattern condition on accuracy or RT on the Non-rigid Bending test,
data from the two conditions were collapsed for further analysis.
Cogn Process (2013) 14:163–173 167
123
than the ‘‘medium’’ trials (M = 1.66 s, SD = .60 s),
t(115) = -8.17, p = .05. Thus, analogous to the findings
in mental rotation that RT increases with increases in the
level of simulated rotation required for the item (Shepard
and Metzler 1971), we find that RT on the Non-rigid
Bending test increases with the amount of simulated
bending required for an item.
Performance across all measures
Due to the relatively high performance on the ‘‘easy’’
(M = .92, SD = .13) and ‘‘medium’’ (M = .85, SD = .13)
trials on the measure, subsequent analyses focused on the
‘‘hard’’ trials (M = .72, SD = .18), as this condition had
the largest range of individual performance and overall
variance. To study the relationship among the four tasks,
Pearson’s correlations were conducted.
A summary of the correlations, means, and standard
deviations for all four of the tasks is shown in Table 1. All
correlations were positive. Using Cohen’s (1988) guide-
lines, accuracy on the Non-rigid Bending test and accuracy
on the Mental Brittle Transformation test were moderately
correlated, r(114) = .381, p \ .001. Accuracy on the Non-
rigid Bending test and accuracy on the Paper Folding test
were also moderately correlated, r(114) = .36, p \ .001,
and accuracy on the Non-rigid Bending test and accuracy
on the Mental Rotation test were weakly correlated,
r(114) = .28, p \ .01.
Based on this administration, the Paper Folding, Mental
Brittle Transformation, and Mental Rotation tests were all
reliable (Chronbach’s alpha statistics were .79, .87, and
.80, respectively). Both conditions of the Non-rigid Bend-
ing test, Gridlines and Texture, were also reliable
(Chronbach’s alpha statistics were .98 and .95,
respectively).3
To further investigate the relations between the spatial
skills used in the Non-rigid Bending test and the spatial
skills used in each of the other measures (Mental Rotation
test, Paper Folding test, and Mental Brittle Transformation
test), a multiple linear regression analysis was conducted to
evaluate how well performance on the other spatial tests
predicted performance on the Non-rigid Bending test. The
predictor variables included performance on the Mental
Rotation test, Paper Folding test, and Mental Brittle
Transformation test. The criterion was performance on the
Non-rigid Bending test. The linear combination of the three
spatial measures significantly predicted performance on the
Non-rigid Bending test, with the combined spatial
measures explaining 22.0 % of the variance, R2 = .22,
F(3,112) = 10.51, p \ .001. Of the three spatial measures
used in the study, only performance on the Mental Brittle
Transformation test (b = .29, p \ .01) and Paper Folding
test (b = .25, p \ .01) significantly predicted Non-rigid
Bending test performance, with performance on the Mental
Brittle Transformation test having the largest effect. Per-
formance on the Mental Rotation test did not predict Non-
rigid Bending test performance, n.s. A summary of the
results from this multiple linear regression is shown in
Table 2.
Because results from the previous multiple linear
regression indicated that performance on the Mental
Rotation test did not significantly contribute to predicting
performance on the Non-rigid Bending test, a stepwise
regression was used to further investigate the relationship
between the four measures in our study. In order to find the
best subset of measures that predict bending performance, a
forward selection stepwise regression was conducted with
performance on the three measures (Mental Brittle Trans-
formation test, Paper Folding test, and Mental Rotation
test) as the predictors, and performance on the Non-rigid
Bending test as the criterion. In this type of regression, the
variable most strongly correlated to the criterion is inserted
first (Thompson 1978). Consistent with the outcome of the
previous regression, only two of the three predictors sig-
nificantly contributed to the model. The combination of
Table 2 Multiple linear regression analyses predicting performance
on Non-rigid Bending test
Predictors b SE b b
MBT .55 .04 .29**
PF .01 .00 .25*
MRT .06 .10 .06
The predictors in this multiple linear regression were average per-
formance on the Mental Brittle Transformation test (MBT), Paper
Folding test (PF), and Mental Rotation test (MRT)
* p \ .01
** p \ .001
Table 1 Summary of correlations, means, and standard deviations
for scores on the Non-rigid Bending test (NRB), Mental Rotation test
(MRT), Paper Folding test (PF), and Mental Brittle Transformation
test (MBT)
Measure NRB MRT PF MBT M SD
1. NRB – .28* .36** .38** .72 .18
2. MRT – – .46** .35** .34 .18
3. PF – .27* 9.17 4.22
4. MBT – – – – .28 .31
* p \ .01
** p \ .001
3 Due to the dichotomous nature of the Non-rigid Bending test, split-
half reliability was also calculated for both Gridlines and Texture
conditions (Spearman-Brown coefficients were .89 and .60,
respectively).
168 Cogn Process (2013) 14:163–173
123
performance on the Mental Brittle Transformation test
(b = .31, p \ .01) and the Paper Folding test (b = .28,
p \ .01) predicted 21.7 % of the variance in Non-rigid
Bending test performance, R2 = .22, F(2,113) = 15.63,
p \ .001, with performance on the Mental Brittle Trans-
formation test still having the largest contribution. Perfor-
mance on the Mental Rotation test did not contribute
significantly to the model, n.s. A summary of the results
from this stepwise regression is shown in Table 3 (top
section).
To further understand the relations between the skills
assessed by each of the measures, three additional forward
selection stepwise regressions were conducted. Each of the
stepwise regressions was used to examine how combina-
tions of the four measures in this study (Mental Rotation
test, Paper Folding test, Mental Brittle Transformation test,
and Non-rigid Bending test) predicted performance on each
of the individual tests. A summary of the results from all of
the stepwise regressions is shown in each section of
Table 3. The Mental Rotation and Non-rigid Bending tests
did not significantly predict performance on each other
after taking account of performance on the Paper Folding
and Mental Brittle Transformation tests. Similarly, per-
formance on the Paper Folding and Mental Brittle Trans-
formation tests was not related once performance on the
other two tests was taken into account.
Discussion
This study investigated the relations among rigid and non-
rigid mental transformations. We developed a new measure
of bending, a skill entailing a continuous non-rigid mental
transformation. Participants’ performance on the Non-rigid
Bending test appeared to be related to the amount of mental
bending required for each item. This finding is analogous to
findings in mental rotation research that performance is
related to the amount of simulated rotation (Shepard and
Metzler 1971). Performance on this new test was predicted
by performance on measures of other non-rigid transfor-
mations (folding and breaking), but not rigid transforma-
tions (rotating). Interestingly, performance on folding and
breaking was both related to rotating and bending, but
breaking and folding were not strongly correlated with
each other. Combining the findings of this study with dis-
tinctions in spatial thinking made in the extant literature
(e.g., Chatterjee 2008; Resnick and Shipley in press;
Shepard and Cooper 1982), we propose that the skills used
for rigid mental transformations are distinct from those
used for non-rigid mental transformations, and skills used
to manipulate intrinsic spatial properties are distinct from
those used for extrinsic spatial properties. These relation-
ships are illustrated in Fig. 3. This figure spatially
represents an interpretation of our results. Skills that share
underlying cognitive processes are shown as partially
overlapping. So, for example, mental bending shares pro-
cesses with both breaking and folding, but not the same
ones, since folding and breaking do not share any resource
that is common to all four visualization skills.
We found a relation among our three measures of non-
rigid mental transformations (breaking, folding, and
bending), and a lack of relation between bending and
rotating. Resnick and Shipley (in press) found a similar
disassociation in experts. Expert geologists and expert
chemists performed equally well on a measure of mental
rotation, but the geologists were superior to the chemists in
performance on a measure of mental breaking. Taken
together this suggests that the spatial skills used for non-
rigid mental transformations are distinct from the spatial
skills used for rigid mental transformations. Thus, we
interpret the vertical position of tests in Fig. 3 as indicating
the degree (or amount) of non-rigid mental simulation
required to solve the tasks. Rigid rotation requires little or
no skill, bending the most, and folding and breaking an
intermediate level because they require local piece-wise
Table 3 Stepwise regression analyses predicting performance on
each spatial measure
Criterion Predictors R2 DR2 b SE b b
NRB Model 1 .15 .15**
MBT .23 .05 .38**
Model 2 .22 .07*
MBT .18 .05 .31*
PF .01 .00 .28*
MRT Model 1 .22 .22**
PF .02 .00 .46**
Model 2 .27 .06*
PF .02 .00 .40**
MBT .15 .05 .25*
PF Model 1 .22 .22**
MRT 10.72 1.92 .46**
Model 2 .27 .06*
MRT 9.09 1.94 .40**
NRB 5.75 1.94 .25*
MBT Model 1 .15 .15**
NRB .64 .15 .38**
Model 2 .21 .07*
NRB .51 .15 .31*
MRT .45 .15 .27*
Measures used in these stepwise regressions include average perfor-
mance on the Non-rigid Bending test (NRB), the Mental Brittle
Transformation test (MBT), Paper Folding test (PF), and the Mental
Rotation test (MRT)
* p \ .01
** p \ .001
Cogn Process (2013) 14:163–173 169
123
rigid transformations. Thus, we suggest that previous
findings of a partial overlap in performance on rotation and
folding tests (see Harris et al. in press) reflect a pair of
underlying cognitive processes necessary to simulate
folding as a partially non-rigid and partially rigid event.
This account suggests the same is true for mental simula-
tion of breaking. Note, in this interpretation, we are not
committing to a particular spacing on this dimension, and
the folding is positioned halfway between bending and
rotating in the diagram for simplicity, in fact the two
intermediate cases maybe closer to one end than the other.
The correlational data do not allow precise relative posi-
tioning at this point.
Although folding and breaking may both require rigid
and non-rigid reasoning processes, they share little vari-
ance. We hypothesize that this reflects a fundamental dif-
ference between the two types of transformations: breaking
results in many separate objects, whereas folding produces
a single object. This distinction matches Chatterjee’s
(2008) intrinsic–extrinsic dichotomy and is represented by
the horizontal axis of the diagram in Fig. 3. Although we
can talk about a broken object as one thing, we may need to
reason about breaking in terms of the extrinsic relations
among the fragments of the original object, and conversely,
paper folding may require going from the segmented
independent parts of an object to a single integrated whole.
By this account, the intermediate position of bending and
rotating on the intrinsic/extrinsic dimension implies that
these visualizations may require a combination of intrinsic
and extrinsic cognitive processes. As noted above, we are
not committed to the particular metric spacing of the
transformations, so bending and rotating likely, on balance,
lean more toward intrinsic processing. That they are not
purely intrinsic is consistent with there being multiple
strategies for approaching these test items. Rotating for
example may be solved with either a holistic strategy that
operates on the single object (e.g., Cooper, 1975; Shepard
and Metzler 1971) or a piece-wise strategy of rotating one
part at a time (e.g., Goksun et al. in press; Just and Car-
penter 1976; Khooshabeh et al. in press).
By considering the nature of natural events and devel-
oping tools that can measure skills in reasoning about a
broad array of mentally simulated events, we have identi-
fied a new type of spatial reasoning and achieved some new
insights regarding previous findings on mental rotation and
folding. The motivation for this study and the context for
the interpretations we have offered reflect an interest in
broadening the scope of spatial cognition research to con-
sider complex spatial problems and the information
humans use to solve them. We refer to this as an ecological
approach to spatial cognition. The ‘‘ecological approach’’
discussed here is analogous to Gibson’s ecological
approach to visual perception (Gibson 1986) in that there is
an emphasis on how the structure of information available
in the environment guides an organism’s actions. Similarly,
we suggest that the deformational information available in
the environment guides the mental transformations and
spatial reasoning skills. We diverge from Gibson in not
advocating a theory of direct perception or action. We want
to emphasize the potential value of an analysis of the
potential structured information available in the ambient
optic array that could inform the observer about past
events, independently from discussion of how the infor-
mation influences behavior. Taking an ecological approach
and studying skills required to visualize transformations
present in the environment should help assure that what is
learned in the laboratory can inform cognitive science
about the many spatial problems humans reason about in
their lives.
Evaluating how well the spatial problems that are used
to study spatial cognition represent the spatial problems
encountered by humans as individuals or a species is dif-
ficult. We advocate an ecological approach to this chal-
lenge, one that focuses on the structure of the information
available to solve spatial problems. Some previous research
has also taken this approach (e.g., Gilden and Proffitt 1989;
Kaiser et al. 1986; Proffitt and Gilden 1989). For example,
Chariker et al. (2011) begin to characterize the spatial
challenges faced by neuroanatomy students as they attempt
to learn to recognize 3D structures from 2D cross sections.
In the study reported here, we focused on a spatial problem
Fig. 3 A Venn Diagram illustrates the shared variance among the
four spatial skills measures. In the diagram, the overlap betweencircles indicates shared variance and suggests a shared set of spatial
skills. No overlap between circles indicates that there was little shared
variance, and possibly no significant overlap in skills. We hypothesize
that the vertical axis represents a continuum of non-rigid to rigid
transformations and that the horizontal axis represents the continuum
of extrinsic to intrinsic spatial skills
170 Cogn Process (2013) 14:163–173
123
faced by geologists when they see layers of rocks that are
bent or curved as a result of a geological transformation—
bending. For the initial foray into this research domain, we
focused on the structured information in geology for sev-
eral reasons: the discipline represents the array of
mechanical, thermal, and chemical spatial relations that
result from objects’ interactions; geologists self-report high
levels of spatial skills across a range of problems on both of
Chatterjee’s (2008) dichotomies (Hegarty et al. 2010); and
the cognitive skills employed by geologists have begun to
be characterized (Kastens and Ishikawa 2006).
To develop a program of research on understanding
events, we look to structural geology, which is the study of
the three-dimensional distribution of rocks and how this
spatial pattern records the history of the rocks. Consider the
events of the earth from the perspective of a structural
geologist. As rocks move relative to other rocks, the
motion may be associated with ductile changes (a contin-
uous change in shape—e.g., bending) or brittle changes
(discontinuous changes—e.g., breaking) in the rocks. In
some regions (e.g., tectonic plate margins), movements
involve a combination of ductile and brittle changes,
whereas in other regions, the motions of the region can be
described as a rigid motion. Thus, geologic events may be
divided into rigid and non-rigid changes. For rigid changes,
all the points in the changing object maintain their spatial
relations to each other. For non-rigid changes, or defor-
mation, which is a change in the shape or dimensions of a
rock resulting from stress or strain, the spatial relations
may change. All deformations, rigid and non-rigid, can be
described by a displacement field, which is a three-
dimensional array of vectors describing the local spatial
changes in a three-dimensional volume. Although defor-
mation may be quite complex, any deformation field can be
decomposed into four basic components: translation, rota-
tion, dilation, and distortion (Malvern 1969). The compo-
nent fields may be summed to recover the original complex
deformation. Thus, all complex non-rigid changes can be
characterized by the sum of four simple spatial changes.
Cognitive science has notably not investigated mental
simulation of all four components of complex non-rigid
changes. As noted, most research has focused on rotation,
and some work has investigated mental translation (e.g.,
Shepard 2001; Shepard and Cooper 1986; Shepard and
Metzler 1971). The non-rigid transformation employed in
this experiment could be constructed from rotations and
translations. The critical distinction between rigid and non-
rigid transformations that we see in our data may reflect
fundamental differences in mentally simulating a single
rotation and mentally simulating transformations that vary
in magnitude across the object or image. Given the lack of
research on mental simulation of dilation and distortion, we
suggest that there is critical need for a research program in
this area that addresses the questions of how the mind
simulates all four different types of changes, how it sim-
ulates spatially varying changes, and how it simulates
combinations of changes. The value of such a program
would be to expand our understanding of the varieties of
transformations people can mentally simulate, and identify
potential constraints. We are not committed to a decom-
positional account where we expect to explain all mental
transformations based on the sum of four basic mental
transformation primitives. To the contrary, we are merely
pointing out that the landscape of what we do not know
about mental transformations seems to be fairly extensive.
A key idea here is that the physical sciences, such as
geology, can provide the requisite description of spatial
information; cognitive scientists may profitably look to
these sciences for help in understanding the categories of
spatial problems that we face in the world. This approach
also identifies experts in the natural sciences as people who
will be particularly good at solving the types of spatial
problems that are central to that science (Lubinski and
Benbow 1992). Thus, an ecological approach to delineating
spatial problems can also provide insight into the types of
skills that should be developed and fostered in the class-
room. Since high spatial ability has been found to predict
success in science disciplines (Shea et al. 2001), identify-
ing and categorizing the spatial information in spatial
problems could guide new approaches to curriculum
development to help train students’ spatial thinking skills,
develop their general problem-solving skills, and enhance
learning in the physical sciences.
We found that measures of mental bending, breaking,
folding, and rotating demonstrates a complex, overlapping
pattern of variance. We interpret this pattern as reflecting
an underlying dimension of spatial reasoning processes that
are used to mentally simulate rigid and non-rigid trans-
formations. Future work supporting our specific interpre-
tations would also support an ecological approach and
illustrate the value of considering the types of spatial
information characterized by the natural sciences—those
sciences charged with describing the world we live in and
must reason about.
Acknowledgments Except the first author, all subsequent authors
are listed alphabetically. This research was supported by a grant to the
Spatial Intelligence and Learning Center, funded by the National
Science Foundation (SBE-0541957 and SBE-1041707), and by a
Fostering Interdisciplinary Research on Education grant, funded by
the National Science Foundation (grant number DRL-1138619).
References
Appelle S (1972) Perception and discrimination as a function of
stimulus orientation: the ‘oblique effect’ in man and animals.
Psychol Bull 78(4):266–278
Cogn Process (2013) 14:163–173 171
123
Caplan PJ, MacPherson GM, Tobin P (1985) Do sex-related
differences in spatial abilities exist? A multilevel critique with
new data. Am Psychol 40(7):786–799
Carroll JB (1993) Human cognitive abilities: a survey of factor-
analytic studies. Cambridge University Press, Cambridge
Chariker JH, Naaz F, Pani JR (2011) Computer-based learning of
neuroanatomy: a longitudinal study of learning, transfer, and
retention. J Educ Psychol 103(1):19–31. doi:10.1037/a0021680
Chatterjee A (2008) The neural organization of spatial thought and
language. Semin Speech Lang 29:226–238
Cohen J (1988) Statistical power analysis for the behavioral sciences,
2nd edn. Erlbaum, Hillsdale
Cooper LA (1975) Mental rotation of random two-dimensional
shapes. Cognit Psychol 7:20–43
Ekstrom RB, French JW, Harman H, Derman D (1976) Kit of factor-
referenced cognitive tests. Educational Testing Service,
Princeton
Gibson JJ (1986) The ecological approach to visual perception.
Lawrence Erlbaum Associates, Hillsdale, New Jersey
Gibson EJ, Spelke ES (1983) The development of perception. In:
Mussen P (series ed), Flavell JH, Markman E (eds) Handbook of
child psychology, vol 3. Wiley, New York
Gibson EJ, Walker AS (1984) Development of knowledge of visual-
tactual affordances of substance. Child Dev 55:453–460
Gibson EJ, Owsley CJ, Johnston J (1978) Perception of invariants by
five-month old infants: differentiation of two types of motion.
Dev Psychol 14(4):407–415
Gilden DL, Proffitt DR (1989) Understanding collision dynamics.
J Exp Psychol 15(2):372–383
Goksun T, Goldin-Meadow S, Newcombe N, Shipley T (in press)
Individual differences in mental rotation: What does gesture tell
us? Cognit Process Spatial Learn Reason Process
Guildford JP, Lacey JL (1947) Printed classification tests, A.A.F. In:
Army Air Force Aviation Psychology Program Research
Reports, No. 5 (pp. 931). U.S. Government Printing Office,
Washington, DC
Harris J, Hirsh-Pasek K, Newcombe N (in press) Understanding
spatial transformations: Similarities and differences between
mental rotation and mental folding. Cognit Process Spatial Learn
Reason Process
Hegarty M, Waller D (2004) A dissociation between mental rotation
and perspective-taking spatial abilities. Intelligence 32:175–191
Hegarty M, Kriz S, Cate C (2003) The roles of mental animations and
external animations in understanding mechanical systems. Cog-
nit Instr 21(4):325–360
Hegarty M, Crookes R, Dara-Abrams D, Shipley TF (2010) Do all
science disciplines rely on spatial abilities? Preliminary evidence
from self-report questionnaires. In: Hoelscher C, Shipley TF,
Bateman J, Olivetti M, Newcombe N (eds) Spatial cognition VII.
Springer, Berlin, pp 85–94
Just MA, Carpenter PA (1976) Eye fixations and cognitive processes.
Cognit Psychol 8:441–480
Kaiser MK, Jonides J, Alexander J (1986) Intuitive reasoning about
abstract and familiar physics problems. Mem Cognit
14(4):308–312
Kastens KA, Ishikawa T (2006) Spatial thinking in the geosciences
and cognitive sciences: a cross-disciplinary look at the intersec-
tion of the two fields. In: Manduca CA, Mogk DW (eds) Earth
and mind: how geologists think and learn about the earth. The
Geological Society of America, Boulder, pp 53–76
Khooshabeh P, Hegarty M, Shipley TF (2012) Individual differences
in mental rotation: piecemeal vs. holistic processing. Exp
Psychol 1:1-8
Kozhevnikov M, Motes MA, Rasch B, Blajenkova O (2006)
Perspective-taking vs. mental rotation transformations and how
they predict spatial navigation performance. Appl Cognit
Psychol 20:397–417
Linn MC, Petersen AC (1985) Emergence and characterization of sex
differences in spatial ability: a meta-analysis. Child Dev
56(6):1479–1498
Lubinski D, Benbow CP (1992) Gender differences in ability and
preferences among the gifted: implications for the math-science
pipe-line. Curr Dir Psychol Sci 1:61–66
Maccoby EE, Jacklin CN (1974) The psychology of sex differences.
Stanford University Press, Stanford
Malvern LE (1969) Introduction to the mechanics of a continuous
medium. Prentice-Hall, Englewood Cliffs
McGee MG (1979) Human spatial abilities: psychometric studies and
environmental, genetic, hormonal, and neurological influences.
Psychol Bull 86:899–918
Mehta Z, Newcombe F (1991) A role for the left hemisphere in spatial
processing. J Devoted Study Nerv Syst Behav 27(2):153–167
Milivojevic B, Johnson BW, Hamm JP, Corballis MC (2003) Non-
identical neural mechanisms for two types of mental transfor-
mation event-related potentials during mental rotation and
mental paper folding. Neuropsychologia 41(10):1345–1356
Newcombe N, Shipley TF (in press) Thinking about Spatial Thinking:
new typology, new assessments. In: Gero JS (ed) Studying visual
and spatial reasoning for design creativity. Springer, New York
Pani JR (1993) Limits on the comprehension of rotational motion:
mental imagery of rotations with oblique components. Percep-
tion 22:785–808
Pani JR, Dupree D (1994) Spatial reference systems in the compre-
hension of rotational motion. Perception 23:929–946
Pani JR, William CT, Shippey G (1995) Determinants of the
perception of rotational motion: orientation of the motion to
the object and to the environment. J Exp Psychol Hum Percept
Perform 21:1441–1456
Pani JR, Jeffres JA, Shippey G, Schwartz K (1996) Imagining
projective transformations: aligned orientations in spatial orga-
nization. Cognit Psychol 31:125–167
Pani JR, William CT, Shippey GT (1998) Orientation in physical
reasoning: determining the edge that would be formed by two
surfaces. J Exp Psychol Hum Percept Perform 24(1):283–300
Pani JR, Chariker JH, Dawson TE, Johnson N (2005) Acquiring new
spatial intuitions: learning to reason about rotations. Cognit
Psychol 51:285–333
Peters M, Laeng B, Latham K, Jacson M, Zaiyona R, Richardson C
(1995) A redrawn Vandenberg and Kuse metal rotation test—
different version and factors that affect performance. Brain
Cognit 28(1):39–58
Pittenger JB, Shaw RE (1975) Aging faces as viscal-elastic events:
implications for a theory of nonrigid shape perception. J Exp
Psychol Hum Percept Perform 1(4):374–382
Proffitt DR, Gilden DL (1989) Understanding natural dynamics. J Exp
Psychol Hum Percept Perform 15(2):384–393
Resnick I, Shipley TF (in press) Breaking new ground in the mind: An
initial study of mental brittle transformation and mental rigid
rotation in science experts. Cognit Process Spatial Learn Reason
Process
Rock I (1973) Orientation and form. Academic, New York
Shea DL, Lubinski D, Benbow CP (2001) Importance of assessing
spatial ability in intellectually talented young adolescents: a
20-year longitudinal study. J Educ Psychol 93:604–614
Shepard RN (2001) Perceptual cognitive universals as reflections of
the world. Behav Brain Sci 24:581–601
Shepard RN, Cooper LA (1986) Mental images and transformations.
MIT Press/Bradford Books, Cambridge
Shepard RN, Metzler J (1971) Mental rotation of three-dimensional
objects. Science 171:701–703
172 Cogn Process (2013) 14:163–173
123
Shipley TF, Zacks JM (eds) (2008) Understanding events: from
perception to action. Oxford University Press, New York
Smith TR, Pellegrino JW, Golledge RG (1982) Computational
process modeling of spatial cognition and behavior 14(4):
305–325
Spelke ES (1982) Perceptual knowledge of objects in infancy. In:
Mehler J, Garrett M, Walker E (eds) Perspectives on mental
representation. Erlbaum, Hillsdale
Thompson ML (1978) Selection of variables in multiple regression:
part I. A review and evaluation. Int Stat Rev 46(1):1–19
Thurstone LL, Thurstone TG (1941) Factorial studies of intelligence.
Psychom Monogr 2:1–38
Vandenberg SG, Kuse AR (1978) Mental rotations: a group test of
three-dimensional spatial visualization. Percept Motor Skills
47:599–604
Vogel JJ, Bowers CA, Vogel DS (2003) Cerebral lateralization of
spatial abilities: a meta-analysis. Brain Cognit 52(2):197–204
Cogn Process (2013) 14:163–173 173
123