Memory & Cognition

1997,25 (3),32/-332

Mental animation in the visuospatial sketchpad:

Evidence from dual-task studies
VALERIE K. SIMS and MARY HEGARTY
University of California, Santa Barbara, California

We used the dual-task paradigm to provide evidence that inferring the motion of a component of a
mechanical system (mental animation) is a spatial visualization process. In two experiments, partici-
pants were asked to solve mental animation problems while simultaneously retaining either a visuo-
spatial working memory load (a configuration of dots in a grid) or a verbal memory load (a list of let-
ters). Both experiments showed that mental animation interferes more with memory for a concurrent
visuospatial load than with memory for a verbal load. Experiment 1 also showed that a visuospatial
working memory load interferes more with mental animation than does a verbal memory load. Fur-
thermore, Experiment 2 showed that mental animation interferes more with a visuospatial memory
load than does a verbal reasoning task that takes approximately the same amount of time.

It has been reported that when the inventor Tesla thought The Experimental Task
of a design for a new device, he first ran a mental simu- We examined the effects of visuospatial interference
lation of the device for several weeks to see which parts on a task that we refer to as the motion verification task.
were most subject to wear (O'Neill, 1944, cited in Shep- In this task, people are shown a diagram ofa mechanical
ard, 1978). This account is anecdotal, but more system- system and are required to infer the motion of one of the
atic analyses of the notebooks of creative scientists sug- system components on the basis of the spatial configu-
gest that visuospatial imagery plays an important role ration of the system and information about how another
in problem-solving tasks such as scientific discovery component is moving. For example, in the sample item
and invention (E. S. Ferguson, 1977; Miller, 1984; Ners- shown in Figure I, people are told to imagine that the
essian, 1992). In addition, some more practical activities, free end of the rope is being pulled down and their task
such as machine assembly and troubleshooting, have is to verify whether the lower pulley will move in the di-
been characterized as involving "envisioning," "running a rection shown by the arrow. At least on face value, this
mental model," or "simulating the behavior of a system task involves the process of mentally simulating the be-
in the mind's eye" (DeKleer & Brown, 1984; Forbus, havior ofparts of a device, as would be required to invent,
Nielsen, & Faltings, 1991; Gentner & Stevens, 1983; assemble, or troubleshoot a device.
Reiger & Grinberg, 1977; Williams, Hollan, & Stevens, Hegarty and Sims (1994) proposed three stages in the
1983). These studies suggest an important function of dy- solution of a motion verification item. The first stage is
namic spatial imagery in reasoning and problem solving. diagram comprehension. At this stage, the solver in-
Most of the studies mentioned above are based on sub- spects the diagram and constructs a static spatial repre-
jective reports. How can we obtain objective evidence sentation of the configuration of the mechanical system.
that reasoning and problem-solving tasks involve spatial The second stage is mental animation. At this stage, the
visualization processes? In this article, we use the dual- solver infers the motion of the component in question
task paradigm (Brooks, 1968) to provide such evidence. from a given motion (e.g., the motion of the pull rope in
That is, we test whether mechanical reasoning impairs a a pulley system or the driver in a gear system). The third
simultaneously held visuospatial working memory load stage is comparison. At this stage, the solver compares
more than it impairs a simultaneously held verbal work- the inferred motion to the motion indicated by an arrow
ing memory load. on the component in question to determine whether the
answer is true or false.
Hegarty (1992) proposed a piecemeal model ofthe men-
tal animation stage. According to this model, people de-
This research was supported by a National Academy of Education
Spencer Fellowship awarded to M.H. We wish to thank Akira Miyake compose the representation of the mechanical system
and two anonymous reviewers for comments on an earlier version of into smaller units, corresponding to the machine compo-
this manuscript and Stephanie Thompson, Amy Horowitz, and David nents and infer the motion of these components one by
Oglesby for research assistance. Correspondence should be addressed one in a sequence corresponding to the causal sequence
to M. Hegarty, Department of Psychology, University of California,
Santa Barbara, CA 93106 (e-mail: hegarty@condor.psych.ucsb.edu)or
of events in the machine's operation. For example, to
V K. Sims, Psychology Department, Cedar Crest College, 100 College infer the motion of the lower pulley in the item in Fig-
Drive, Allentown, PA 18104 (e-mail: vksims@ccc-s.cedarcrest.edu). ure I, a solver first infers that the rope is moving to the

321 Copyright 1997 Psychonomic Society, Inc.

322 SIMS AND HEGARTY

ofvisual images (Kosslyn, Brunn, Cave, & Wallach, 1985;

Poltrock & Agnoli, 1986; Smith, 1964). Therefore, if
mental animation involves dynamic spatial imagery, then
people who have good spatial visualization ability
should perform well on mental animation tasks. In a pre-
vious study (Hegarty & Sims, 1994), we examined the
relation of spatial visualization ability to accuracy on the
motion verification task. Spatial visualization ability had
strong and consistent effects on performance of the mo-
tion verification task, whereas verbal ability was unre-
lated to performance on this task. Although not conclu-
sive, these results suggest that mental animation is a
spatial visualization process.
Second, a number of recent studies have argued for a
visuospatial inference process on the basis of hand ges-
tures that people make while they are inferring the mo-
tion of parts of mechanical systems (Clement, 1994;
Figure 1. A sample motion verification item. Hegarty & Ferguson, 1993; Narayanan, Suwa, & Mo-
toda, 1994; Schwartz & Black, 1996b). For example,
Schwartz and Black (1996b) have shown that when
right over the upper pulley, then infers that the upper pul- asked to determine the motions of gears in a gear chain,
ley is turning clockwise and that the right rope strand is people begin by enacting the motions of the gears with
moving up. From this information, he/she goes on to their hands and formulate rules of mechanical reasoning
infer that the rope is moving to the right under the lower on the basis of these enactments. In other experimental
pulley, which enables him/her to infer that the lower tasks, such hand movements or enactments have been in-
pulley is turning counterclockwise. According to this terpreted as an externalization of spatial imagery trans-
model, people have to make a longer chain of inferences formations (Reisberg & Logie, 1993).
when inferring the motion of components later in the A third source of evidence for a spatial visualization
causal chain of events in the operation of a mechanical process comes from studies measuring the time to imag-
system. Consistent with the model, participants in pre- ine a displacement of a mechanical component as a func-
vious experiments made more errors and required more tion of the size ofthe displacement. For example, Schwartz
time to infer the motion of these components. Further- and Black (1996a) found that the time to infer the rota-
more, eye-fixation data indicated that when asked to tion of two interlocking gears was proportional to the
infer the motion of a particular component, participants angle of rotation of the two gears. This result suggests
typically looked at that component and earlier compo- that inferring gear motion is an analog imagery process.
nents in the causal chain of events, but not at later com- In this article we use the dual-task paradigm (Brooks,
ponents (Hegarty, 1992; Hegarty & Sims, 1994). 1968) to argue for the use of spatial visualization pro-
cesses in mental animation. Selective interference stud-
Spatial Visualization and Mental Animation ies offer particularly convincing evidence for the involve-
The model of mental animation proposed by Hegarty ment of visual or spatial imagery in task performance.
(1992) was intended primarily as an account of how a These studies are based on the logic that if a primary task
person decomposes the task of inferring the motion of and a concurrent visuospatial secondary task interfere
components of a complex mechanical system made up with each other, then visuospatial processes are used in
of several interacting components. The model was ex- performing the primary task. It must also be shown that
pressed as a production system. However, it was not the primary task and a concurrent verbal secondary task
meant to imply that the process of inferring the motion interfere with each other to a lesser extent.
of a component of a mechanical system is a rule-based In the past, dual-task studies have provided evidence
reasoning process. In other words, evidence for a piece- for separate verbal and spatial stores in working memory
meal mental animation process does not inform us (Baddeley, 1986; Baddeley & Lieberman, 1980). Specif-
whether individuals infer the motion at each "link" in the ically, Baddeley has proposed that working memory con-
causal chain by matching production rules against a sists of a central executive and at least two "slave sys-
propositional representation of a mechanical system (a tems" specialized for representing different types of
structural description) or by a spatial visualization pro- information. These slave systems include the articu-
cess in which they mentally transform an image of one latory loop, which is specialized for storing and pro-
or more components of the mechanical system. cessing verbal information, and the visuospatial sketch-
Several recent studies suggest that at least early in pad, which is specialized for spatial storage and informa-
practice, mental animation might involve dynamic spa- tion processing (Baddeley, 1986). Theories of working
tial imagery. Previous research has suggested that spatial memory differ regarding how much information pro-
visualization ability requires the accurate transformation cessing is attributed to these slave systems. For example,
 MENTAL ANIMATION 323

Logie (1995) has suggested that the slave systems merely There is also a possibility that time to recall the memory
store and rehearse verbal and spatial representations, re- load will be slowed due to mental animation, making it
spectively, and that more complex processing of these less accessible in memory. Thus, we argue that either de-
representations is carried out by the central executive. creased accuracy or slower performance on either mo-
In contrast, Shah and Miyake (1996) have argued for sep- tion verification or the secondary task can be taken as
arate pools of working memory resources involved in evidence that the two tasks compete for the same work-
complex spatial information processing (e.g., mental ing memory resources. Note that previous studies using
rotation) and verbal information processing (e.g., lan- the dual-task paradigm have not always examined all
guage comprehension). The present research is based on possible dependent measures. For example, one task that
the assumption that the visuospatial sketchpad is spe- has been used extensively to measure verbal working
cialized for at least the storage and maintenance of spa- memory resources is the reading span task (Daneman &
tial representations. This research also provides infor- Carpenter, 1980; Just & Carpenter, 1992). This task re-
mation on the locus of more complex spatial information quires subjects to read a group of sentences and recall
processing. the final word of each sentence. The measure of reading
The dual-task paradigm has also been applied to the span is the number of words recalled. That is, it is a mea-
study of higher level cognitive processes. For example, sure of accuracy in recalling a memory load, but it does
Baddeley (1992) used this paradigm to demonstrate that not measure performance on the primary task (sentence
playing chess involves the working memory resources of comprehension).
the visuospatial sketchpad. Similarly, Kruley, Sciama, Whether impairment occurs on either the primary or
and Glenberg (1994) used dual-task studies to demon- secondary task probably depends on the participants'
strate that comprehension of text accompanied by pic- judgments of which task is more important. In a previous
tures involves the visuospatial sketchpad. dual-task study in which mental rotation was the primary
We describe two experiments that examined the role task, Shah and Miyake (1996) found that the interference
of spatial working memory in motion verification. In was evident primarily in loss of the concurrent memory
Experiment 1 we contrasted the interference effects that load (i.e., the secondary task). Similarly, in a dual-task
occur when motion verification is paired with a visuo- study in which comprehension of text accompanied by
spatial working memory load to the interference effects pictures was the primary task, Kruley et al. (1994) found
that occur when motion verification is paired with a ver- interference effects only on the secondary task, despite
bal memory load. In Experiment 2, we contrasted the in- extensive efforts to emphasize accuracy on the sec-
terference effects that occur when visuospatial and ver- ondary task. Thus it is possible that participants will
bal memory loads are paired with motion verification to judge any complex spatial task to be more important
the interference effects that occur when each of these than a concurrent memory load task and will allocate
memory loads is paired with a verbal reasoning task that more resources to that task, in which case we would ex-
takes approximately the same amount of time as the mo- pect interference effects to be evident primarily in per-
tion verification task. formance on the secondary task.
We hypothesize that mental animation relies on the
EXPERIMENT 1 storage and processing functions of the visuospatial
sketchpad. Therefore, we expect to find more impair-
In Experiment 1, we examined the degree to which the ment when the dot memory task is paired with motion
motion verification task and two different concurrent verification than when the letter memory task is paired
memory tasks interfere with each other. We will refer to with mental animation. Notice that we argue for greater
the motion verification as the primary task and to the interference by a spatial secondary task than by a verbal
concurrent memory tasks as secondary tasks. In the sec- secondary task, and not for the absence of interference
ondary tasks, participants were asked to retain in memory by a verbal secondary task. Because the verbal secondary
either a spatial array of dots or a list of letters. Previous task was presented visually, it may also require the re-
literature (Kruley et al., 1994; Phillips & Christie, 1977) sources of the visuospatial sketchpad and therefore be
has suggested that retaining a spatial array of dots uses interfered with somewhat by the motion verification
the working memory resources of the visuospatial sketch- task. Furthermore, the task-decomposition and goal-
pad. In contrast, memory for a list of letters has been management components of the motion verification task
shown to use the resources of the articulatory loop (Bad- are most likely to involve the resources of the central ex-
deley & Hitch, 1974). ecutive, so we would expect motion verification to in-
We measured the degree to which each concurrent terfere somewhat with any secondary task.
task interferes with performance on the motion verifica-
tion task and vice versa. If mental animation competes Method
for working memory resources with a memory load, there Participants
are several possible outcomes: The memory load will be Forty-two students from the University of California, Santa Bar-
lost, performance on mental animation will be impaired, bara, took part in this experiment. Two students' data were not in-
or performance on mental animation will be slowed down. cluded in the analyses because they performed lower than chance
324 SIMS AND HEGARTY

level on the motion verification task. Of the remaining 40 partici- A. VlBuo-Spatial Secondary Task
pants, 20 took part in the verbal memory load condition and 20 took
part in the spatial memory load condition. Participants either re-
ceived course credit or were paid $5.00 for their participation. True
Trial:
Design
This study used a 2 (type of secondary task) X 2 (interference)
X 3 (position in causal chain) mixed design. Type of secondary task
was manipulated between subjects. In addition to performing mo-
tion verification, participants were asked to hold a working mem-
ory load that was either verbal or spatial. The presence or absence
of a working memory load was manipulated within subjects. Each False
participant was asked to perform the motion verification task both Trial:
with and without a memory load as interference. For interference
trials, the participant was given a memory load and asked to verify
it after performing motion verification. For the no-interference tri-
als, the participant was given a memory load, was asked to verify B. Verbal Secondary Task
it, and then performed motion verification. Position in the causal
True
chain was also a within-subjects factor. Participants were given mo-
Trial: LVYQMX Q
tion verification trials that asked them to verify one of three posi-
tions in the causal chain of events: beginning, middle, or end.
Four dependent variables were computed: percent error on the
motion verification task, percent error on the secondary task, reac- False
Trial: DMKXRG w
tion time (in seconds) on the motion verification task, and reaction
time (in seconds) on the secondary task. Figure 2. Examples oftrue and false trials for the visuospatial
and verbal secondary tasks.
Apparatus
Stimuli were presented on Macintosh IIci computers using
MindLab (Mcike, Bharucha, Baird, & Stoeckig, 1988) software. grid (false trial). The direction of movement of this dot was deter-
MindLab is a programming shell that presents stimuli and collects mined using a random-number generator and was constrained in
responses. It has a temporal resolution of 16.8 msec. such a way that a straight line was not formed. Examples of true and
false trials are shown in Figure 2A.
Materials Verbal secondary task. The verbal secondary task was adapted
Motion verification task. The motion verification task was from Sternberg (1969) and is similar to verbal interference tasks
used in previous research on mental animation (Hegarty, 1992; Heg- used in other dual-task studies (e.g., Baddeley & Hitch, 1974). On
arty & Sims, 1994). It consisted of24 unique stimuli. Each stimu- each trial, a list of six letters appeared for 5 sec. Letter lists were
lus showed a diagram of a pulley system drawn in black on a white constructed using a random-number generator. No letter was re-
background. In addition, a red box was drawn around one of the peated within a letter list and no vowels were used, thus minimiz-
pulleys, and a red arrow was drawn, showing a possible motion of ing the chances of creating a pronounceable string of letters. For
this pulley. The participant's task was to decide if the arrow showed verification, participants were shown a single letter, and they had to
the correct direction of movement if the free end of the rope in the decide if this letter was a part of the initial list of six letters. Trials
pulley system were pulled. Twelve stimuli showed true movements, were constructed so that half were true and half were false. Exam-
and 12 showed false movements. Figure I shows a sample stimulus. ples of true and false trials are shown in Figure 28.
Participants were presented with diagrams of two different pul- Combination of tasks. The verbal and visuospatial secondary
ley systems and the mirror images ofthese systems (producing left tasks were combined with the motion verification task to produce
and right isomers of the two systems). Each pulley system con- four different conditions. In the visuospatial memory load condi-
tained three pulleys. One pulley was at the beginning of the causal tion, participants were presented with a dot grid as a memory load
chain in the system, the second was in the middle of the causal and were then asked to solve a motion verification problem; finally
chain, and the third was at the end of the causal chain. In Figure I, they were asked to verify the initial memory load. For the verbal
the upper pulley is toward the beginning of the causal chain, be- memory load condition, participants were shown a list of letters,
cause its motion is caused directly by pulling the free end of the pull solved a motion verification problem, and then verified a single let-
rope. In tum, the middle pulley is in the middle of the causal chain ter. In the no-load conditions, participants were presented with ei-
because its motion is caused by the movement of the upper rope, ther a visuospatial or a verbal load, were asked to verify it, and then
which in turn has been caused by pulling on the free end of the rope. solved a motion verification problem. True and false motion veri-
Similarly, the lower pulley is toward the end of the chain of events fication questions were paired with true and false secondary trials,
because its movement takes place as a result of several previous producing equal numbers of the four possible trial types. Addition-
movements, including those of the upper and middle pulleys. ally, trials were constructed so that each pulley system was paired
Visuospatial secondary task. The visuospatial interference with the same number of true and false distractor trials. For both of
task was adapted from Kruley et al. (1994). On each trial, the par- the secondary tasks, as well as the motion verification task, partic-
ticipant was first shown a 4 X 4 grid containing three dots. His/her ipants responded "true" by pressing the "k" key on the computer
task was to memorize the pattern of dots for later verification. The keyboard, and "false" by pressing the "d" key. For ease in respond-
positions of the dots were determined using a random-number gen- ing, a green sticker, labeled "T," was placed on the true key, and a
erator. The only constraint was that the three dots could not fall in red sticker, labeled "F," on the false key.
a straight line. The initial presentation of the dot pattern lasted two
thirds of a second. For verification, participants were presented Procedure
with a second grid containing three dots. For half of the trials, this Participants were run in groups of 1-5, and each group was as-
grid was identical to the initial grid (true trial). For the other half of signed to either the verbal interference or the spatial interference
the trials, one randomly chosen dot was moved by one space on the condition. First. participants were given directions and eight prac-
 MENTAL ANIMATION 325

tice trials for the secondary task. Directions informed them that 20
they would be briefly seeing a series ofletters or a collection of dots
and that they would later have to verify what they had seen. Exam-
ples ofbotb true and false trials were also presented, along with ex- 15
•0
Memory Load
No Memory Load
planations of the correct responses. Next, participants were given
10
directions explaining the motion verification trials. The directions
described all of the parts of the pulley systems and explained the ~
motion verification task. Participants then completed eight practice ...
~
l:l
10

trials for the motion verification task. At this point, participants had III
a chance to ask questions about either task. Participants were then ~
III
given the memory load and no-load versions of the main task. The l:loo
5
order of these tasks was counterbalanced so that approximately half
of the participants received the memory load condition first, and
approximately half received the no-load condition first. At the end
o
of the experiment, participants were thanked and dismissed.
Visuo-Spatial Verbal
Results and Discussion Type of Interference
We examined selective interference effects on four Figure 4. Percent error on the secondary tasks in Experi-
different dependent measures-accuracy on the primary ment 1.
and secondary tasks and reaction time on the primary
and secondary tasks. Using the Dunn-Bonferroni proce- than performance on the motion verification task in the
dure for multiple a priori comparisons, we set the alpha presence of a verbal memory load. We analyzed error on
levelatp < .0125 (Winer, Brown, & Michels, 1991). All re- the motion verification task in a 2 (type of secondary
ported effects are significant at this level, unless other- task) X 2 (presence of a memory load) mixed analysis
wise noted. of variance (ANOYA). As shown in Figure 3, the inter-
We first compare overall performance in the load and action of presence of a memory load with type of sec-
no-load conditions for the motion verification task and ondary tasks was marginally significant [F( 1,38) =
the memory load tasks. We later investigate how these 5.42, MS e = 33.67,p = .025]. Planned comparisons in-
effects are modulated by difficulty of the motion verifi- dicated that when a memory load was present, the spa-
cation items (as measured by distance of the component tial interference group made more errors than did the
in question from the beginning of the causal chain). verbal interference group [t(38) = 3.25, one-tailed].
However, when there was no memory load, the perfor-
Performance on the Motion Verification Task mance of two groups did not significantly differ [t(38) =
We predicted that the visuospatial secondary task 0.97]. These results indicate that motion verification
would interfere more with motion verification than performance was impaired when there was a visuospatial
would the verbal secondary task. If this is the case, per- memory load, as would be expected if both tasks were
formance on the motion verification task in the presence sharing the resources of the visuospatial sketchpad. The
of a visuospatial working memory load should be poorer results also indicate that the groups were matched on
mental animation performance when there was no load.
There was also a marginal main effect for motion verifi-
20-r-------------------, cation performance to be poorer for the group with the
visuospatial secondary task than for the group with the
• Memory Load verbal secondary task [F(l,38) = 5.33, MSe = 132.36,
o No Memory Load p = .026]. The main effect for presence of a memory
load was not statistically significant.
A separate ANOYA on the reaction times for the mo-
tion verification task revealed no significant effects of
either type of interference task or presence of a memory
load on this measure.

Performance on the Secondary Tasks

First, we established that the two secondary tasks were
of equivalent difficulty when a motion verification trial
did not intervene between presentation and verification
of the memory load. In the no-load conditions, the error
Visuo-Spatial Verbal rate for the dot memory task was low (4.17%) and was
not significantly different from the error rate for the let-
Type of Interference ter memory task, which was 3.96% [t(38) = 0.12]. Fig-
Figure 3. Percent error on the motion verification task in ure 4 shows the means and standard errors for error rate
Experiment 1. on the secondary tasks in the load and no-load conditions.
326 SIMS AND HEGARTY

25~-------------r5 Effects of Position in the Causal Chain

on Performance
20
As in previous research (Hegarty, 1992; Hegarty &
Sims, 1994), participants made more errors in verifying
the motion of components later in the causal chain [see
Figure 5; F(2,78) = 20.62, MSe = 109.21]. Reaction
time to verify the motion of a pulley also increased mono-
tonically with its distance from the beginning of the
causal chain of events [F(2,78) = 76.10, MSe = .847].
Therefore, items about components later in the causal
chain are more difficult and require more time. According
to the piecemeal model of mental animation (Hegarty,
o o 1992), this is because components are mentally animated
Beginning Middle End one by one in the order of the causal chain of events.
Position in the Causal Chain To investigate how difficulty and processing time of a
Figure 5. Percent error and reaction time on the motion verifi- motion verification item affect loss ofa concurrent mem-
cation task as a function of position in the causal chain in Exper- ory load, we analyzed the effects ofposition in the causal
iment 1. The bars show the percent error, and the line shows the chain on errors and reaction time on the secondary task.
reaction time. This analysis was based on difference scores between the
load and no-load conditions (i.e., the loss of a memory
load when motion verification intervenes between stor-
We predicted that memory for the secondary task age and recall minus the loss of memory load when there
would be poorer in the load conditions than in the no- is no intervening task). These difference scores are shown
load conditions and that this decrement in performance in Figure 6. The difference scores were then subjected to
would be greater for the visuospatial task (dot memory) a 3 (position in the causal chain) X 2 (type of secondary
than for the verbal task (letter memory). A 2 (type of task) ANaYA. In the analysis based on errors, the effect
secondary task) X 2 (presence of a memory load) mixed of position in the causal chain did not reach significance
ANaYA revealed a significant main effect of presence [F(2,76) = 1.99, MSe = 135.83], nor did its interaction
ofamemory load [F(l,38) = 33.36, MS e = 34.67]. More with type of interference task [F(2,76) = 2.19, MSe =
critically, presence of a memory load interacted with type 135.83].
of secondary task [F(l,38) = 9.52, MSe = 34.67]. As In the analysis based on reaction time, there was a sig-
predicted, there was a greater decrement in performance nificant effect of position in the causal chain [F(2,76) =
on memory for dot configuration (an accuracy difference 5.31, MSe = .21], indicating that verification of both
ofl1.66%; SD= 9.99) than on memory for letters (3.54%; memory loads was slowed down as a function ofthe time
SD = 6.25). Again, this is consistent with the view that taken for the intervening motion verification item. Posi-
mental animation and dot memory compete for the re- tion in the causal chain did not interact with type of sec-
sources of the visuospatial sketchpad. The main effect ondary task.
for a higher error rate by the group with the visuospatial
memory task also approached statistical significance
[F(l,38) = 5.2, MSe = 70.12,p = .02]. 20~------------...,

A comparison of reaction times in the no-load condi- • Visuo-Spatlal

tions indicated that it took longer to verify a letter (M = o Verbal
1.44 sec, SD = .32) than a dot pattern [M = .96 sec,
SD = .24, t(38) = 5.38]. Therefore, we standardized the
reaction time data by converting the data for each task to
z scores. These z scores were then subjected to a 2 (type
of interference task) X 2 (presence of a memory load)
mixed ANayA. Participants were slower on the sec-
ondary task when a motion verification trial intervened
between presentation and verification of the memory
load (M = .46, SD = 1.11) than when a motion verifi-
cation trial did not intervene [M = - .44, SD .== .59, o
F(l,38) = 37.58, MSe = .43]. Reaction time was not af- Beginning Middle End
fected significantly by type of secondary task or the Position in the Causal Chain
interaction of type of secondary task with presence of a Figure 6. Difference in percent error on the secondary task be-
memory load. tween the load and no-load conditions in Experiment 1.
 MENTAL ANIMATION 327

EXPERIMENT 2 Design
This study used a 2 (type of reasoning task) X 2 (type ofmem-
Experiment 1 established that mental animation inter- ory interference) mixed design. Type ofreasoning task was manip-
ulated within subjects. Each participant performed two primary
feres more with memory for the spatial configuration of
tasks: the Baddeley and Hitch (1974) verbal reasoning task and mo-
dots than with memory for a sequence of letters. It also tion verification. Type of interference was manipulated between
suggested that the dot memory task interferes more with subjects. Participants completed both main tasks while carrying ei-
mental animation than does the letter memory task. ther a verbal memory load (letters) or a spatial memory load (dot
These results are highly suggestive that mental anima- configuration).
tion uses the resources of the visuospatial sketchpad. As in Experiment I, the motion verification task contained trials
However, we need to address the alternative hypothesis asking about the beginning, middle, and end of the causal chain.
The verbal reasoning task required that participants verify both ac-
that holding a dot configuration in memory consumes tive and passive statements. We measured the same four dependent
more working memory resources in general than does variables as those in Experiment I.
holding a sequence of letters in memory. If this is the
case, we should observe greater interference when dot Apparatus
memory is paired with any task, and not just when it is As in Experiment I, stimuli were presented on Macintosh IIci
paired with a spatial visualization task. In order to rule computers using Mindlab (Meike et aI., 1988) software.
out this alternative hypothesis, it is necessary to establish Materials
that there is a task with which dot memory does not in- In this experiment, participants completed a verbal reasoning
terfere as much as does letter memory. task and a motion verification task, each while performing a sec-
In Experiment 2, we examined the degree to which the ondary task. The motion verification task and both of the secondary
dot memory task and the letter memory task interfere interference tasks (dot memory and letter memory) were identical
with two different primary tasks: motion verification and to those used in Experiment I.
Verbal reasoning task. The verbal reasoning task used in this
a verbal reasoning task. The verbal reasoning task was
experiment was adapted from Baddeley and Hitch (1974). On each
adapted from Baddeley and Hitch (1974) and was chosen trial, a sentence appeared followed by two letters. Each sentence
because pilot studies established that it takes approxi- described the ordering of those letters; for example:
mately the same amount of time as the motion verifica-
A precedes B AB
tion task. In this task, the participant is asked to verify
the order of two letters of the alphabet. Researchers have The letters presented after the sentence were always in alphabetic
argued that this task primarily requires the resources of order. The participant's task was to decide whether the sentence ac-
curately described the order of the letters. Twenty-four trials were
the central executive, with some dependence on the ar- constructed in the following manner. Sentences were written either
ticulatory loop, but that it does not use the resources of in the active voice (as in the above example) or the passive voice
the visuospatial sketchpad. Thus, performance on this (e.g., "A is preceded by B AB.") and used either the verb precedes
task is impaired somewhat by articulatory suppression, or the verb follows. Twelve sentences asked about the ordering of
but not at all by spatial tapping (Farmer, Berman, & the letters a and b, 12 about the ordering of c and d, and 12 about
Fletcher, 1986; Hitch & Baddeley, 1976). We predict an the ordering of e and f Half of the trials were true, and half were
false. As in the motion verification condition, the secondary verbal
interaction between type of memory load and type of
and visuospatial tasks were paired with the verbal reasoning task so
main task (motion verification or verbal reasoning task). that equal numbers of true and false distractor trials were paired
First, participants should show greater impairment when with active and passive statements in the verbal reasoning task.
the dot memory task is paired with motion verification Combination of tasks. The verbal and visuospatial secondary
than when it is paired with the verbal reasoning task. tasks were combined with motion verification and verbal reasoning
Second, participants should show more impairment to produce four different conditions. In the visuospatial memory
when the letter memory task is paired with the verbal load conditions, participants were presented with a dot grid as a
memory load, then solved a motion verification problem or a
reasoning task than when the letter memory task is verbal reasoning problem, and finally verified the initial memory
paired with motion verification. load. In the verbal memory load conditions, participants were
shown a list of letters, solved a motion verification problem or a
verbal reasoning problem, and then verified that a single letter was
Method a part of the memory load. As in Experiment I, true and false mo-
Participants tion verification questions were paired with true and false sec-
Forty-six students from the University of California, Santa Bar- ondary trials, producing equal numbers of the four possible trial
bara, received course credit to take part in Experiment 2. Six of types. Again, participants responded by pressing "k" for true and
these participants were deleted from the final analysis because their "d" for false.
performance on the verbal reasoning task was poorer than chance.
Subsequent analysis of their answers revealed that they had an- Procedure
swered fewer than a quarter of the passive items correctly, suggest- Participants were run in groups of 1-5, and all performed both
ing that they had misunderstood the instructions for the verbal rea- motion verification and verbal reasoning as primary tasks. Half of
soning task. Participants were assigned to one of two groups: the participants performed these tasks with a visuospatial memory
visuospatial interference or verbal interference. The final sample load, and half performed these tasks with a verbal memory load.
contained 20 participants in the visuospatial interference group, The order of tasks was counterbalanced so that approximately half
and 20 participants in the verbal interference group. of the participants received the motion verification task first and
328 SIMS AND HEGARTY

approximately half received the verbal reasoning task first. As in (13.6%) than did participants who received the verbal
Experiment I, participants received instructions and eight practice memory load (12.1%).
trials for each experimental task before they performed that task. In
each condition of the experiment, participants received two repli-
cations of the stimuli.
Performance on the Secondary Tasks
Wepredicted that participants would show greater im-
Results pairment on dot memory when it was paired with the
motion verification task than when it was paired with the
The data were analyzed separately for error on the pri- verbal reasoning task and that participants should show
mary tasks, reaction time on the primary tasks, error on more impairment on the letter memory task when it was
the secondary tasks, and reaction time on the secondary paired with the verbal reasoning task than when it was
tasks. The alpha level was again set atp < .0125. We first paired with the motion verification task. We analyzed er-
compare performance on the primary tasks (verbal rea- rors on the secondary task in a 2 (type of primary task)
soning and motion verification) and then examine how X 2 (type of secondary task) ANOYA. As shown in Fig-
these tasks affected verification of the memory load (ei- ure 7, there was a significant interaction between type of
ther visuospatial or verbal). We later investigate how primary task and type of secondary task [F(l,3 8) =
these effects were modulated by difficulty of the motion 20.54, MSe = 61.01]. As predicted, participants were
verification items (position in the causal chain) and the more impaired on dot memory when they performed an
verbal reasoning items (verification of active or passive intervening motion verification item and were more im-
sentences). paired on letter memory when they performed an inter-
veningverbal reasoning item. As in Experiment 1,planned
Performance on the Primary Tasks comparisons indicated that mental animation interfered
We analyzed performance on the primary tasks in two more with the visuospatial secondary task than with the
2 (type of primary task) X 2 (type of secondary task) verbal secondary task [t(38) = 4.05]. However, verbal
mixed ANOVAS, with errors and reaction times as the reasoning did not interfere significantly more with the
dependent variables. Performance on the motion verifi- verbal secondary task than with the visuospatial sec-
cation and verbal reasoning tasks revealed that as in our ondary task. Furthermore, dot memory was more impaired
pilot testing, the mean time taken for a motion verifica- by motion verification than by verbal reasoning [t( 19) =
tion trial (3.22 sec, SD = 1.19)was approximately equal to 3.39] and letter memory was more impaired by verbal
the mean time taken for a verbal reasoning item [3.46 sec, reasoning task than by motion verification [t(19) = 3.15].
SD = 1.11, F(1,38) = 1.09. MS e = 1.06]. However, there These results provide strong evidence that mental ani-
was a nonsignificant trend for participants to make more mation and dot memory share working memory re-
errors on the verbal reasoning task (12.87%, SD = 13.1) sources (i.e., the visuospatial sketchpad) that are distinct
than on the motion verification task [8.18%, SD = 8.29, from those used by verbal reasoning and letter memory.
F(1,38) = 3.75,MSe = 117.13,p = .06]. Since participants were slower to verify letters (M =
A possible account of the results of Experiment 1 is 1.81 sec, SD = .32 sec) than dot configurations (M =
that holding a dot configuration in memory consumes 1.37 sec, SD = .23 sec), as in Experiment 1, the reaction
more working memory resources in general than does time data were converted to z scores. We then analyzed
holding a sequence of letters in memory. If this is the the standardized data in a 2 (type of primary task) X 2
case, performance on both the motion verification and (type of secondary task) ANOVA. The interaction of
verbal reasoning tasks should be poorer when they are
paired with dot memory than when they are paired with
letter memory. However, in Experiment 2, performance 30.---------------,
on the primary tasks was not significantly affected by the Motion VeriOcation Trials
type of secondary tasks (dot memory or letter memory) 25 Verba1 Reasoning Trials
[F(1,38) = .55, MSe = 128.46 for error rate; F(1,38) =
1.10, MSe = 1.61 for reaction time]. &.. 20

~
This experiment did not replicate the trend in Experi-
ment 1 for mental animation performance to be more im- _ 15

paired by a visuospatial working memory load than by a =

verbal working memory load. There were no significant
interactions of type of primary task with type of sec-
j 10

ondary task [F(1 ,38) < 1 for both error rate and reaction
time]. The results were in the same direction as those in
Experiment 1. Participants receiving the visuospatial in- o
terference task made more errors on the motion verifi- Visuo-Spatiai Verbal
cation task (9.2%) than did participants receiving the Type of Interference
verbal memory load (7.1 %). However, these participants Figure 7. Percent error on the secondary tasks in Experi-
also made more errors on the verbal reasoning task ment 2.
 MENTAL ANIMATION 329

type of primary task with type of secondary task was not 40,---------------,
VIsuo-Spatial
statistically significant[F(l ,38) = 2.21, MSe = .63], nor
VemaJ
was the main effect oftype of secondary task [F(l,38) =
1.37]. Although the primary tasks slowed down the ver- 30

ification of the memory loads in the predicted direction,

the significant interference effects in Experiment 2 were
in loss of the memory loads.
J 20

Effects of Position in the Causal Chain on

Performance
1
l l.
10

For the motion verification trials, error and reaction

time data were analyzed in a 3 (position in the causal
o
c~ain) X 2 (type of secondary task) ANaYA. As in pre-
Beginning Middle End
VIOUS research (Hegarty, 1992; Hegarty & Sims, 1994)
and in Experiment 1, participants made more errors in Position in the Causal Chain
verifying the motion of components later in the causal Figure 9. Percent error on the distractor task in Experiment 2
chain [see Figure 8; F(2,76) = 21.83, MSe = 127.12]. as a function of position in the causal chain.
Reaction time to verify the motion of a pulley also in-
c~eased monotonically with its distance from the begin-
rung of the causal chain of events [see Figure 8; F(2,76) = explanation ofthis effect is that the middle pulley in both
62.71, MS e = 1.42]. Position in the causal chain did not pulley systems has more connections to other compo-
interact with type of secondary task for the measures of nents in the system, making it more visually complex.
performance on the motion verification task. Therefore, more working memory resources might be re-
To investigate how difficulty and processing time of a quired to represent this pulley. However, this effect
motion verification item affect loss of a concurrent might also be an artifact due to more difficult distractor
memory load, we analyzed the effects of position in the trials being paired with items about the middle pulley
causal chain on errors and reaction time for the sec- (aUparticipants received the same pairings of distractors
ondary task in a 3 (position in the causal chain) X 2 with motion verification items).
(ty~e.ofs.econdary task) ANaYA. As shown in Figure 9,
In the analysis of reaction times for the secondary
position III the causal chain interacted with type of sec- task, there was also a marginal interaction of type of sec-
ondary task [F(2,76) = 15.87, MS e = 56.54]. Simple ef- ondary task with position in the causal chain [F(2,76) =
fects indicated that position in the causal chain affected 3.32, MSe = .714,p = .04]. Simple effects analyses in-
memory for a dot pattern [F(2,76) = 22.26, MSe = 56.54] dicated a marginal effect of position in the causal chain
more than it affected memory for a list ofletters [F(2,76) for the visuospatial secondary task [F(2,76) = 4.16,
= 2.84, MSe = 56.54]. Curiously, motion verification MSe = .714, P = .02]. As Figure 10 shows, reaction time
items about the middle pulley in the causal chain caused on the visuospatial secondary task increased with dis-
more loss of the visuospatial working memory load than tance of a component from the beginning of the causal
did items about the beginning or end pulleys. A tentative chain. In contrast, there was no significant effect of po-
sition in the causal chain on reaction time for the verbal
secondary task (F < I). This result is consistent with the
25 ....- - - - - - - - - - - - - - ,
interpretation that mental animation and dot memory
share the resources of the visuospatial sketchpad-if a
trial involves more mental animation steps, it renders the
20
dot pattern less accessible in working memory. An alter-
native explanation is that the increased time to verify a
dot pattern is merely a function of the amount of time in-
tervening between first seeing the pattern and later ver-
ifying it. However, note that this effect does not occur for
.......... the verbal memory load.

Effects of Voice (Passive or Active) in the Verbal

Reasoning Task
As in previous research (Baddeley & Hitch, 1974), there
Beginning Middle End was a trend for participants to make more errors verify-
Position in the Causal Chain ing passive sentences (16.04%, SD = 20.1) than active
sentences [9.69%, SD = 10.77, F(l,38) = 4.78, MS e =
Figure 8. Percent error and reaction time on the motion verifi-
~ation task as a function of position in the causal chain in Exper-
1.69.02,p < .05]. They also took more time to verify pas-
tment 2. The bars show the percent error, and the line shows the srve sentences (3.81 sec, SD = 1.36) than active sentences
reaction time. [3.18 sec,SD = .97,F(I,38) = 36.06,MSe = .22]. How-
330 SIMS AND HEGARTY

Vlsuo-Spatlal
erent in the visuospatial sketchpad. In these situations,
Verbal
"mental model" refers to a spatial representation of the
referent of the text and/or picture. In research on rea-
soning about physical systems, the term "mental model"
has been used to refer to a dynamic spatial representation
that allows one to mentally simulate the operation of a
system (DeKleer & Brown, 1984; Forbus et aI., 1991;
Gentner & Stevens, 1983; Reiger & Grinberg, 1977;
Williams et aI., 1983). Our research provides evidence
that this type of dynamic mental model also requires the
resources of the visuospatial sketchpad.
The study of dynamic spatial reasoning tasks such as
mental animation can provide important information
Beginning Middle End
about the structure of working memory. For example, a
Position in the Causal Chain point of current controversy is whether the visuospatial
Figure 10. Reaction time on the distractor task in Experiment 2 sketchpad has specialized resources for both storage and
as a function of position in the causal chain. processing of spatial information or whether it is merely
a static store for spatial information, with spatial pro-
cessing being carried out by the central executive (Logie,
ever,the difficulty ofa verbal reasoning task (as measured 1995; Shah & Miyake, 1996). It would be interesting to
by voice) affected neither the accuracy nor the time to compare the memory load requirements of mental ani-
verify a memory load (F < 1 for both measures). Diffi- mation and a task that merely requires inspecting a sta-
culty also did not interact with type of secondary task. tic representation. For example, Hegarty (1992) com-
Note that passive sentences did not increase the reaction pared the motion verification task used in this research
time to verify a dot pattern. Thus the observed effects of with a static task in which subjects had to verify state-
position in the causal chain on reaction time to verify a ments about the spatial relations between components in
dot pattern can be interpreted as selective interference a pulley system diagram. If the visuospatial sketchpad is
effects and not merely decay with time required by the specialized for both storage and processing, dynamic
intervening (primary) task. mental models should require more spatial working
memory resources than do static spatial representations.
GENERAL DISCUSSION The present experiments provided preliminary evi-
dence in support ofthe view that there are working mem-
In this research, we demonstrated that mental anima- ory resources specialized for dynamic spatial process-
tion of a pulley system interferes more with memory for ing. First, our research showed a dissociation between a
a concurrent visuospatial array than with memory for a dynamic spatial reasoning task and a verbal reasoning
list of letters. Experiment 1 also showed that a visuo- task that has been shown to be highly dependent on the
spatial working memory load interferes more with men- central executive (Farmer et aI., 1986; Hitch & Baddeley,
tal animation than does a verbal memory load. Further- 1976). Second, more dynamic tasks (i.e., motion verifi-
more, Experiment 2 indicated that mental animation cation trials that involved more spatial transformations)
interferes more with a visuospatial memory load than tended to interfere more with memory for a stored spa-
does a verbal reasoning task that takes approximately the tial array.
same amount of time. These results argue for a dissoci- In the dual-task paradigm, impairment on either the
ation between the working memory resources required primary or the secondary task is evidence that two tasks
by mental animation and those required by verbal rea- share the same resources. Although there was evidence
soning and suggest that mental animation is dependent for impairment in both of our experiments, there was
on the resources of the visuospatial sketchpad. some inconsistency between experiments as to which
The evidence presented in this article is consistent task was impaired. In both experiments, the secondary
with the model proposed by Baddeley (1986), who sug- task was impaired. This finding is in line with other
gested that working memory contains separate "slave dual-task studies in which the primary task was a rela-
systems" for storing and maintaining visuospatial and tively complex spatial task (Kruley et aI., 1994; Shah &
verbal information. Our results also add to a growing Miyake, 1996). In Experiment 1, the primary task also
body of evidence that the visuospatial sketchpad is in- tended to be impaired, but this trend was not replicated
volved in high-level comprehension and reasoning tasks in Experiment 2. Although the subjects in the two ex-
that involve spatial representations (Baddeley, 1992; periments were given the same instructions, a compari-
Kruley et aI., 1994). In fact, the visuospatial sketchpad son across experiments suggests that they weighted the
has been proposed as the site of mental models. Kruley tasks differently. When dot memory was paired with
et al. have suggested that when people read a text ac- mental animation, the mean error on the dot memory
companied by a picture (e.g., a description of a spatial task was lower in Experiment I (15.8%) than in Experi-
environment), they construct a mental model of the ref- ment 2 (23.3%), and the mean error for mental anima-
 MENTAL ANIMATION 331

tion was higher in Experiment 1 (12.9%) than in Exper- DEKLEER, J., & BROWN. J. S. (1984). A qualitative physics based on
iment 2 (9.3%). Therefore, subjects in Experiment 1 confluences. Artificial Intelligence. 24. 7-83.
FARMER. E. w.. BERMAN, J. V. E, & FLETCHER. Y. L. (1986). Evidence
seem to have allocated relatively more resources to the for a visuo-spatial scratchpad in working memory. Quarterly Journal
secondary task. Further research is required to study the ofExperimental Psychology. 38A, 675-688.
influences on the relative weighting given to the two FERGUSON. E. L., & HEGARTY, M. (1994). Properties of cognitive maps
tasks in a dual-task study. constructed from text. Memory & Cognition. 22,455-473.
In this experiment, we used mental animation of pul- FERGUSON. E. S. (1977). The mind's eye: Nonverbal thought in tech-
nology. Science. 197.827-836.
ley systems as an example of a dynamic mechanical rea- FORBUS, K. D.. NIELSEN, P.• & FALTINGS, B. (1991). Qualitative spatial
soning task. Other research has shown that like pulley reasoning-The clock project. Artificial Intelligence. 51,417-471.
systems, gear, lever, and more complex systems are also GENTNER, D., & STEVENS, A. L. (EDS.) (1983). Mental models. Hills-
animated piecemeal, and that ability to mentally animate dale, NJ: Erlbaum.
HEGARTY. M. (1992). Mental animation: Inferring motion from static
these systems is related to spatial visualization ability
displays of mechanical systems. Journal of Experimental Psychol-
(Hegarty & Steinhoff, 1994). Therefore, we would expect ogy: Learning. Memory. & Cognition. 18, 1084-1102.
the selective interference results observed here to gener- HEGARTY, M., & FERGUSON, J. M. (1993, November). Strategy change
alize to the mental simulation ofother mechanical devices. . with practice in a mental animation task. Paper presented at the an-
Finally, we might ask whether inferring the motion of nual meeting of the Psychonomic Society, Washington, DC.
HEGARTY. M., & SIMS. V. K. (1994). Individual differences in mental an-
a mechanical device from a static diagram (mental ani- imation during mechanical reasoning. Memory & Cognition, 22,4 I 1-
mation) is a mental imagery process. Several sources of 430.
evidence converge on this view. First, mental animation HEGARTY, M., & STEINHOFF. K. (1997).Individual differences in use of
draws on the resources of the visuospatial sketchpad, diagrams as external memory in mechanical reasoning. Learning &
Individual Differences. 9. 19-42.
which has been characterized as being specialized for the
HITCH. G. J.• & BADDELEY. A. D. (1976). Verbal reasoning and work-
storage and processing of mental images (Logie & Bad- ing memory. Quarterly Journal ofExperimental Psychology. 28, 603-
deley, 1990; Riesberg & Logie, 1993). Second, we have 621.
provided evidence elsewhere (Hegarty & Sims, 1994) that JUST. M. A., & CARPENTER, P. A. (1992). A capacity theory of compre-
mental animation performance is correlated with spatial hension: Individual differences in working memory. Psychological
Review. 99, 122-149.
visualization ability, which in turn has been character- KOSSLYN. S. M., BRUNN.J.• CAVE. K. R..& WALLACH. R. W. (1985). In-
ized as involving the storage and processing of mental dividual differences in mental imagery: A computational analysis. In
images (Kosslyn et al., 1985; Poltrock & Agnoli, 1986; S. Pinker (Ed.), Visual cognition (pp. 195-243). Cambridge, MA:
Smith, 1964). Third, Schwartz and Black (l996a) found MIT Press.
KRULEY, P., SCIAMA, S. c., & GLENBERG, A. M. (1994). On-line pro-
that the time to mentally simulate the rotation of two
cessing of textual illustrations in the visuospatial sketchpad: Evi-
gears is proportional to the angle of rotation, suggesting dence from dual-task studies. Memory & Cognition 22, 261-272.
an analog imagery process. Finally, when people reason LOGIE. R. H. (1995). Visuo-spatial working memory. Hove, U'K: Erl-
about dynamic physical systems, they frequently make baum.
hand gestures that simulate the behavior of the different LOGIE, R. H.. & BADDELEY, A. D. (1990). Imagery and working mem-
ory. In P. 1. Hampson, D. F. Marks, & 1. T. E. Richardson (Eds.), Im-
device components (Clement, 1994; Hegarty & Fergu- agery: Current developments (pp. 103-128). London: Routledge.
son, 1993; Narayanan et al., 1994; Schwartz & Black, MEIKE, B., BHARUCHA.J. J.• BAIRD, J. C; & STOECKIG. K. (1988). Mind-
1996b). These gestures are seen as an externalization of Lab. Santa Barbara. CA: Intellimation.
internal imagery processes. These different sources of MILLER. A. I. (1984). Imagery in scientific thought. Boston: Birkhauser.
NARAYANAN, N. H.• SUWA. M., & MOTODA, H. (1994). A study of dia-
evidence point to a central role of dynamic spatial im-
grammatic reasoning from verbal and gestural data. Proceedings
agery in mechanical reasoning. of the Annual Meeting of the Cognitive Science Society. 16.652-
657.
REFERENCES NERSESSIAN, N. (1992). How do scientists think? Capturing the dynam-
ics of conceptual change in science. In R. Giere (Ed.), Cognitive mod-
BADDELEY. A. D. (1986). Working memory. Oxford: Oxford University els of science (pp. 3-44). Minneapolis: University of Minnesota
Press. Press.
BADDELEY. A. D. (J 992). Is working memory working? The fifteenth O·NEILL. J. J. (1944). Prodigal genius: The life of Nikola Tesla. New
Bartlett lecture. Quarterly Journal of Experimental Psychology, 44. York: Ives Washburn.
1-31. PHILLIPS. W. A.• & CHRISTIE. D. F. M. (1977). Interference with visu-
BADDELEY. A. D .• & HITCH. G. (1974). Working memory. In G. H. alization. Quarterly Journal of Experimental Psychology, 29, 637-
Bower (Ed.), The psychology oflearning and motivation (pp. 47-89). 650.
San Diego: Academic Press. POLTROCK, S. E.• & AGNOLI, F. (1986). Are spatial visualization ability
BADDELEY. A. D.• & LIEBERMAN. K. (1980). Spatial working memory. and visual imagery ability equivalent? In R. 1. Sternberg (Ed.), Ad-
In R. Nickerson (Ed.), Attention and performance XVIII (pp. 521- vances in the psychology ofhuman intelligence (pp. 255-296). Hills-
539). San Diego: Academic Press. dale, NJ: Erlbaum.
BROOKS. L. R. (1968). Spatial and verbal components of the act of re- REIGER. c.. & GRINBERG, M. (1977). The declarative representation and
call. Canadian Journal ofPsychology, 22,349-368. procedural simulation of causality in physical mechanisms. Pro-
CLEMENT, J. (1994). Imagistic simulation and physical intuition in ex- ceedings ofthe International Conference on Artificial Intelligence, 5.
pert problem solving. Proceedings of the Annual Conference of the 250-255.
Cognitive Science Society. 16, 201-206. RIESBERG. D.. & LOGIE, R. (1993). The ins and outs of working mem-
DANEMAN. M .. & CARPENTER. P. A. (1980). Individual differences in ory: Overcoming the limits on learning from imagery. In B. Roskos-
working memory and reading. Journal of Verhal Learning & Verbal Ewoldson, M. 1. Intons-Peterson, & R. E. Anderson (Eds.).lmagery,
Behavior. 19.450-466. creativity. and discovery (pp. 39-70). Amsterdam: North-Holland.
332 SIMS AND HEGARTY

SCHWARTZ, D. L., & BLACK, 1. B. (I 996a). Analog imagery in mental SMITH, I. M. (1964). Spatial ability: Its educational and social signifi-
model reasoning: Depictive models. Cognitive Psychology. 30, 154- cance. London: University of London Press.
219. STERNBERG, S. (1969). Memory scanning: Mental processes revealed
SCHWARTZ, D. L., & BLACK, J. B. (1996b). Shuttling between depictive by reaction time experiments. American Scientist, 57, 421-457.
models and abstract rules: Induction and fall-back. Cognitive Sci- WILLIAMS, M. D., HOLLAN, J. D., & STEVENS, A. L. (1983). Human rea-
ence, 20, 457-497. soning about a simple physical system. In D. Gentner & A. L. Stevens
SHAH, P., & MIYAKE, A. (1996). The separability of working memory (Eds.), Mental models (pp. 131-153). Hillsdale, NJ: Erlbaum.
resources for spatial thinking and language processing: An individ- WINER, B. J., BROWN, D. R., & MICHAELS, K. M. (1991). Statistical pro-
ual differences approach. Journal ofExperimental Psychology: Gen- cedures in experimental design (3rd ed). New York: McGraw-HilI.
eral, 125,4-27.
SHEPARD, R. N. (1978). Externalization of mental images and the act of
creation. In B. S. Randhava & W. E. Coffman (Eds.), Visual learning,
thinking, and communication (pp. 133-189). San Diego: Academic (Manuscript received November 14, 1994;
Press. revision accepted for publication September 22, 1995.)