Rhesus monkeys manipulate mental images

Cognition 228 (2022) 105225

Available online 14 July 2022

Thomas C. Hassett

, Victoria K. Lord

, Robert R. Hampton

Department of Psychology, Emory University, Atlanta, GA, United States of America

Emory National Primate Research Center, Atlanta, GA, United States of America

ARTICLE INFO

Keywords:

Mental imagery

Mental rotation

Nonhuman primate

ABSTRACT

Humans form mental images and manipulate them in ways that mirror physical transformations of objects.

Studies of nonhuman animals will inform our understanding of the evolution and distribution among species of

mental imagery. Across three experiments, we found mostly converging evidence that rhesus monkeys formed

and rotated mental images. In Experiment 1, monkeys discriminated rotations of sample images from mirror

images, and showed longer response latencies with greater rotation as is characteristic of human mental rotation.

In Experiment 2 monkeys used a rotation cue that indicated how far to mentally rotate sample images before

tests, indicating a precision of better than 30

◦

in discriminating rotations. Experiment 3 yielded mixed evidence

on whether the rotation cue shortened decision times as has been found in humans. These results show that

rhesus monkeys manipulate mental images.

Humans simulate the physical world with mental images, as when

we visualize re-arranging furniture before physically moving it (Moul-

ton & Kosslyn, 2009; Pearson & Kosslyn, 2015). Human mental images

are characterized by at least three properties: 1) they partially recapit-

ulate perception without concurrent perceptual input, 2) they are

accessible to introspection, and 3) they can be manipulated in ways that

parallel the physics of the objects mentally represented (Ganis,

Thompson, & Kosslyn, 2004; Kosslyn, 1980, 1988; Neiworth & Rilling,

1987; Shepard & Cooper, 1982). Mental imagery may also be an

important cognitive tool for other animals, particularly given the fact

that other animals cannot use linguistic representation as do humans

(Grifn, 1976; Premack, 1983).

Nonhumans have been reported to demonstrate two of the three

characteristics of mental images described above. First, they clearly

mentally represent some features of visual stimuli, partially recapitu-

lating perception as is the case in humans. This is shown by the ability to

accurately match to sample and do other kinds of visual cognitive tasks

(e.g., Adachi, Kuwahata, & Fujita, 2007; Basile & Hampton, 2013;

auer & Belger, 2018; Miller, Erickson, & Desimone, 1996; Neiworth &

Rilling, 1987). Second, two lines of evidence indicate that monkeys

introspect about mental representations, a capacity sometimes known as

phenomenal vision. Metacognition experiments show that monkeys

know when they remember an image, indicating introspection (Basile,

Schroeder, Brown, Templer, & Hampton, 2014; Hampton, 2001;

Zakrzewski, Johnson, & Smith, 2017). “Blindsight” experiments show

that monkeys, like humans, manifest at least two modes of visual

perception, an implicit one and a potentially explicit mode (Andersen,

Basile, & Hampton, 2014; Ben-Haim et al., 2021; Cowey & Stoerig,

1995; Moore, Rodman, & Gross, 2001). We address the third property of

mental imagery here: that mental transformations parallel physical

transformations.

Studies of human mental rotation were important in the “cognitive

revolution,” during which mental events gained credibility in scientic

explanations of behavior. Researchers found that rotating a mental

image shared properties with rotating a physical object. For example, it

takes twice as long both to rotate a steering wheel 180

◦

than 90

◦

, and to

visualize doing so. Participants in these studies decided whether a

rotated shape was identical to, or a mirror image of, an upright com-

parison shape (Premack, 1983; Shepard & Metzler, 1971). The farther

the shape was rotated, the longer it took to decide (Shepard & Cooper,

1982; Shepard & Metzler, 1971).

There have been comparatively few studies of mental rotation

involving nonhuman animal subjects, and comparisons of the published

work does not reveal clear relationships with ecological or phylogenetic

variables. The rst studies were done with pigeons, and these birds

exhibited “rotational invariance.” They were more accurate than ex-

pected by chance across rotations, but reaction time did not correlate

with rotation (Hollard & Delius, 1982). Rotational invariance suggests

* Corresponding author at: Emory National Primate Research Center: Laboratory of Comparative Primate Cognition, 954 Gatewood Rd, Atlanta, GA 30322, United

States of America.

E-mail address: [email protected] (T.C. Hassett).

Contents lists available at ScienceDirect

Cognition

journal homepage: www.elsevier.com/locate/cognit

https://doi.org/10.1016/j.cognition.2022.105225

Received 14 June 2021; Received in revised form 2 July 2022; Accepted 8 July 2022

Cognition 228 (2022) 105225

that pigeons may not rely on manipulation of mental images to identify

rotated objects. In a related but rather different test, pigeons were found

to accurately anticipate the position of a temporarily occluded moving

clock hand, but this is explained as readily by timing as by mental

rotation (Neiworth & Rilling, 1987). Given that pigeons encounter ob-

jects in rapidly varying orientations while ying, they might have

evolved processes that supports rapid recognition of objects in varying

orientations without the time-consuming process of mental rotation

(Hollard & Delius, 1982; Lohmann, Delius, Hollard, & Friesel, 1988).

However, it could also be argued that other species, such as aquatic

animals and arboreal primates could benet similarly from rapid

recognition of rotated objects and they have either been reported not to

show rotational invariance, or show mixed patterns of behavior. A test of

mental rotation in a single sea lion revealed evidence of mental rotation

(Mauck & Dehnhardt, 1997). A study involving a single lion-tailed

macaque, a largely arboreal species, reported the monkey erred more

and was slower on rotation trials compared to non-rotation trials, but

performance did not systematically change with angular rotation (Bur-

mann, Dehnhardt, & Mauck, 2005).

Studies of mental rotation in terrestrial nonhuman primates report

similarly inconsistent results. A study of Guinea baboons found evidence

of mental rotation when stimuli were presented to the right visual eld,

however the ndings did not replicate in another study by the same

research group (Hopkins, Fagot, & Vauclair, 1993; Vauclair, Fagot, &

Hopkins, 1993). Two studies involving rhesus monkeys reported either

no evidence of mental rotation (Nekovarova, Nedvidek, Klement,

Rokyta, & Bures, 2013), or mixed results of mental rotation where one of

three monkeys showed the characteristic increase in response latency

with angular rotation (K

ohler, Hoffmann, Dehnhardt, & Mauck, 2005).

In another study involving ve rhesus monkeys, researchers reported a

systematic decrease in accuracy as angular rotation increased, however

reaction time was not reported (Parr & Heintz, 2008).

Some of the differences reported in the studies above may be due to

differences in amount of training in the tasks used in these tests, or in

prior experience subjects had manipulating objects in their lifetimes.

Human children improve greatly in mental rotation tasks in the rst ve

years of life (Frick, Hansen, & Newcombe, 2013), consistent with the

possibility that experience and practice play a critical role (Newcombe,

2002). Here we provide a relatively large group of monkeys with

extensive training to determine whether or not they are capable of

mental rotation.

The variation in ndings on mental rotation across species may be

due to additional factors in addition to differences in training and

experience. One interesting potential source of variation is cognitive

differences that might associate with differences in behavioral ecology.

For instance, the nding of rotational invariance in pigeons described

above may reect the need for pigeons to especially rapidly recognize

rotated scenes and objects (Hollard & Delius, 1982). However, meth-

odological differences between publications are also a likely cause of

differences in ndings. It is notable that few studies of nonhumans have

used methods consistent with those detailed by Shepard and Metzler

(1971) and subsequent human mental rotation studies. Requiring sub-

jects to discriminate rotated images from mirror-images prevents sub-

jects from using local features of stimuli to solve the task, and response

latency is a critical outcome variable for assessing the correspondence

between physical and mental rotation (Cohen & Kubovy, 1993; Delius &

Hollard, 1995; Shepard & Cooper, 1982). Use of mirror-image stimuli

and/or response latency have not always been reported in nonhuman

studies (Nekovarova et al., 2013; Parr & Heintz, 2008). Comparing

ndings across species would be facilitated by a more standardized

approach using these core features of the work done with humans. In

this study, we developed a monkey analog of the mental rotation task

using a delayed match-to-sample procedure with mirror image

discrimination stimuli and both latency and accuracy as outcome

measures.

In Experiment 1 monkeys demonstrated the hallmark of mental

rotation, taking longer to nd a matching shape the more the shape was

rotated. In Experiment 2 monkeys used a cue that indicated how far to

mentally rotate an image while the image was out of sight. In Experi-

ment 3 we tested whether this rotation cue would speed identication of

rotated images using the procedure from Experiment 1. Findings from

Experiment 3 were mixed.

1. Experiment 1

Monkeys were trained in a mental rotation task that closely paral-

leled methods used with humans (Fig. 1A, Cooper & Shepard, 1973;

Vauclair et al., 1993). Monkeys studied two-dimensional shapes that

disappeared after they touched them. After 500 ms the studied shape

and its mirror image appeared, both rotated up to 120

◦

. Monkeys were

rewarded for selecting the rotated version of the studied shape, avoiding

the mirror image. Critically, mirror images contain identical geometric

features, making it impossible to identify the rotated match by features

alone (Cooper & Shepard, 1973; Shepard & Metzler, 1971). We hy-

pothesized that if monkeys mentally rotate images, and such rotation

takes time, as it would if done physically, then they would take longer to

nd the correct shape the more it was rotated (Shepard & Cooper, 1982).

1.1. Subjects

We tested six adult male rhesus monkeys (mean age: 9.8 years old)

that had extensive experience with cognitive testing and match-to-

sample tasks. Monkeys had ad libitum access to water and received

their daily caloric intake through a combination of nutritionally

balanced reinforcement pellets and monkey biscuits. They were indi-

vidually housed due to social incompatibility but had visual and acoustic

contact with conspecics. Procedures complied with U.S. law, the Na-

tional Research Council guide for the care and use of laboratory animals,

and were approved by the Emory University Institutional Animal Care

and Use Committee.

1.2. Apparatus

Computerized testing systems were mounted on monkeys’ home

cages and consisted of a touch-sensitive LCD monitor (Elo TouchSys-

tems, Menlo Park, Ca), two food dispensers (Med Associates Inc., St.

Albans, VT), and were controlled by custom programs written in Visual

Studio 2013 (Microsoft Corporation).

1.3. Procedure

Monkeys performed a match-to-sample task with mirror image dis-

tractors (Fig. 1A). Trials were initiated by touching a green start square

twice. A sample shape then appeared centrally on screen. Samples were

drawn from a pool of 10 stimuli: 5 shapes and their mirror images

(Fig. S1). After touching the sample shape twice, the screen went blank

for 500 ms. At test, the sample shape and its mirror image were pre-

sented side by side, both rotated either 0

◦

, 30

◦

, 60

◦

, 90

◦

, or 120

◦

. The

position of the target and distractor were counterbalanced across trials.

Selections of the rotated sample shape were followed by a reinforcing

food pellet, positive audio feedback, and a 3-s inter-trial interval. Se-

lections of the mirror image shape were followed by negative audio

feedback and a 6-s time-out. All 10 stimuli appeared once as the correct

response at each angle, randomly distributed in each 50-trial session.

Testing concluded for each monkey when it had completed at least 20

sessions and accuracy was signicantly above chance at all orientations

simultaneously over the last two blocks of ve sessions (signicance at p

< .05 as determined by binomial tests conducted for each orientation

individually).

T.C. Hassett et al.

Cognition 228 (2022) 105225

1.4. Results and discussion

Analyses of reaction time here and throughout used correct trials

only. Across all conditions, monkeys averaged 79% correct, meaning

that on average 21% of trials were not included in the analysis of re-

action time because monkeys chose incorrectly on these trials. Monkeys

took longer to respond on trials with greater rotation, matching the

signature of mental rotation in humans (Fig. 2, upper panel; rmANOVA:

(4, 20)

= 26.551, p < .001,

= 0.841; Shepard & Cooper, 1982; She-

pard & Metzler, 1971). While this result is a clear parallel with human

mental rotation, our monkeys differed from typical results from humans

in that accuracy was lower for larger rotations (Fig. 2, lower panel;

rmANOVA: F

(4, 20)

= 7.822, p = .001,

= 0.605). Longer response times

and greater transformation may have caused forgetting or distortion of

mental images, resulting in the decrease in accuracy found with greater

rotation. A decrease in accuracy is not inconsistent with mental rotation,

and we mitigated the possible effect of differences in accuracy on latency

by only analyzing latencies from correct trials. Nonetheless, because

longer latencies often accompany guessing (e.g., Hampton, 2009), the

increase in latency could plausibly result from monkeys guessing more

with larger rotations, rather than from the time-consuming process of

mental rotation.

To address the potential confounding of response latency and accu-

racy in Experiment 1, we used accuracy instead of latency as the primary

basis for inferring mental rotation in Experiment 2. This task does not

directly test for mental rotation, in the sense of temporally extended

incremental change of a mental representation by rotation around an

axis. However, success in this task does require accurate transformation

of a mental representation while preserving the delity of the corre-

spondence between the mental representation and the represented

image. In Experiment 2, monkeys discriminated identical shapes that

differed only in orientation, ruling out discrimination on any basis other

than orientation.

2. Experiment 2

Monkeys were cued to mentally rotate a sample shape to a specic

orientation. At test they saw two identical shapes; the target was rotated

the cued amount, and the distractor was rotated 30

◦

more or less than

the target (Fig. 1B). The cue specied the amount the target would be

rotated at test, but monkeys were not shown the target rotating. If

monkeys rotate mental images, they should rotate them on cue to

identify the target.

2.1. Subjects & apparatus

The same subjects, stimuli, and apparatus used in Experiment 1 were

used in Experiment 2.

2.2. Procedure

Monkeys learned in stages that a cue indicated the extent to which a

target image would be rotated from the sample orientation.

2.2.1. Initial rotation discrimination training

After monkeys initiated a trial, a sample shape appeared surrounded

by a rotational cue (Fig. 1B). Monkeys touched the sample shape causing

the shape and the cue to rotate clockwise 0

◦

, 30

◦

, 60

◦

, 90

◦

, or 120

◦

. All

rotations took the same amount of time, and involved the same number

of animation frames, with each frame involving a larger movement for

larger rotations. For example, on a 30

◦

trial, the shape and cue were

erased and drawn 30 times, rotating clockwise by 1

◦

each time; on a

120

◦

trial, they were also erased and drawn 30 times, but rotated

clockwise by 4

◦

each time. The duration of the rotation animation was

approximately 950 milliseconds across all rotations, with a standard

deviation of 65 milliseconds. The cue and sample remained onscreen for

500 ms after rotation was complete. The test appeared after the screen

was blank for 500 ms. Both test choices were the previously seen shape.

The target was rotated to the cued orientation, and the distractor was

rotated +/− 60

◦

relative to the target. The left-right position of the

target on the screen, and whether the distractor was rotated more or less

than the target, was counterbalanced across trials. Each target orienta-

tion was used with each shape twice per session in a randomized

sequence, once with a distractor at − 60

◦

and once with a distractor at

+60

◦

, yielding sessions of 100 trials. After monkeys were signicantly

above chance simultaneously at each orientation (signicance at p < .05

as determined by binomial tests conducted for each orientation every 5

sessions) the difference between the targets and distractors was reduced

to 30

◦

. After again meeting criterion, monkeys progressed to the next

stage of training.

Fig. 1. Procedures used to test mental imagery in monkeys.

(A) Mirror image discrimination (Experiment 1): Monkeys

started trials by touching the green square twice. After

touching the sample shape, the screen was blank for 500 ms,

followed by a mirror image discrimination test. (B) Cued angle

discrimination (Experiment 2): The sample shape was touched,

it disappeared, and the cue rotated. Monkeys were rewarded

for selecting the image that was rotated to the extent indicated

by the cue. (C) Cued mirror image and orientation discrimi-

nation (Experiment 3 Cued Phase): Following cue rotation,

monkeys made mirror image and orientation discriminations in

pseudorandom order. In nal testing, monkeys repeated

Experiment 1 for comparison. (For interpretation of the refer-

ences to colour in this gure legend, the reader is referred to the web version of this article.)

Fig. 2. Monkeys took longer to respond and were less accurate with more

rotation. (Top) Median response latency(s) on correct trials; (Bottom) propor-

tion correct.

T.C. Hassett et al.

Cognition 228 (2022) 105225

2.2.2. Sample shape fading

Monkeys learned to follow the rotational cue in the absence of the

target image rotating inside it. We removed the sample shape before the

cue had completed rotating. The point at which the target shape dis-

appeared was pseudorandomly determined from among 10 possible

points between 67% and 100% of the total rotation achieved by the cue.

Monkeys again had to reach criterion. This process was repeated with

the sample shape disappearing between 33% and 67% and nally be-

tween 0% and 33% of cue rotation.

2.2.3. Cue only tests

The sample shape disappeared immediately after being touched, and

the cue then rotated to the to-be-tested orientation without the sample

shape. Monkeys completed this phase when they met the same accuracy

criterion used in training.

2.3. Results and discussion

Monkeys learned to discriminate the shapes at the cued orientation

from distractors rotated +/− 30

◦

(p ≤ .05 by binomial tests at each

orientation; see Table 1). Monkeys therefore mentally rotated the sam-

ple shape to the cued orientation with a delity better than 30

◦

, even

though the orientation of the target shape was indicated only by the cue

and the shape itself did not visibly rotate. Monkeys were also signi-

cantly slower to respond when the amount of rotation dictated by the

cue was large (F(4, 16) = 4.518, p = .012). These results indicate mental

imagery by showing that monkeys formed detailed shape representa-

tions and mentally transformed them with delity to match stimuli

presented on the computer screen. These results show that monkeys are

capable of rotating mental images on cue, but because the monkeys

required extensive training to attain this level of performance, we should

be cautious inferring that monkeys would do this in nature, or even that

such opportunities would arise naturally. These results are important

because Experiment 1 and Experiment 2 used different approaches and

dependent measures yet converged on the conclusion that monkeys

rotate mental images.

One alternative interpretation of the ndings from Experiment 2 is

that monkeys used the orientation of the horizontal or vertical sides of

the cue to guide their choice at test. The upright and horizontal faces of

the rotational cue and the target shape align when oriented at the same

angle (Fig. 1B). A monkey might solve this task by matching the vertical

or horizontal faces of the cue at the time the cue disappeared, with the

faces of the choice shapes at test. If monkeys used this strategy, accuracy

should not vary as a function of the extent of cue rotation because the

difculty of matching the orientation of the cue at offset with the test

images is the same regardless of rotational distance. By contrast, the

more a mental image is transformed, the more likely it is to be forgotten

or distorted, leading to more errors with larger rotations. Monkeys were

signicantly less accurate on trials with greater cue rotation, consistent

with mental imagery and mental rotation (main effect of cued angle: F

(4,

16)

= 11.509, p < .001; see Table 1 for individual data), but not the

alignment hypothesis. We further evaluated the alignment hypothesis in

Experiment 3. We required monkeys to discriminate images that did not

differ in alignment with the rotational cue, but that again differed in the

mirror image dimension, as in Experiment 1. Success in the task in

Experiment 3 cannot be achieved using alignment.

3. Experiment 3

When humans are cued about the orientation in which test shapes

will appear, they rotate their mental images to the correct orientation

before tests, eliminating the longer response times found with larger

rotations (Cooper, 1976; Cooper & Shepard, 1973; Suchan, Botko,

Gizewski, Forsting, & Daum, 2006). In Experiment 3, monkeys repeated

the mirror image discrimination task from Experiment 1, but were either

cued, or not cued, about the orientation in which the test stimuli would

appear. We compared response times on cued mirror image tests

(Fig. 1C) to those from a block of uncued trials from Experiment 1 and a

block of uncued trials run at the end of Experiment 3 (Fig. 1A). If

monkeys mentally rotated images when cued, then they, like humans,

should show less increase in response times with rotation on cued trials

compared to uncued trials. Critically, monkeys could not select the

correct image in the mirror image discrimination tests by using the cue

alone, as they did in Experiment 2 because the task was mirror image

discrimination and both the target and the distractor appeared at the

same orientation.

3.1. Subjects & apparatus

The six monkeys and the materials from Experiment 2 were used

again.

3.2. Procedure

Monkeys rst completed a block of trials using the procedure from

Experiment 2, except two types of tests were pseudorandomly inter-

mixed (Fig. 1C). Half of trials were mirror image discrimination trials,

the other half of trials were orientation discrimination trials, yielding a

total of 200 trials per session. Intermixing the two types of trials ensured

that monkeys continued to use the rotational cue, because half of the

time they needed it to discriminate on the basis of orientation, and the

correct orientation could only be known from the cue. Monkeys were

required to perform signicantly above chance across all angles for 2

consecutive 5-session blocks (signicance at p < .05 as determined by

binomial tests conducted for each orientation every 5 sessions). This

performance criterion was applied to orientation discrimination trials

only, and performance on mirror image trials did not inuence whether

or not a monkey met criterion. Upon meeting this criterion, we planned

for monkeys to proceed to the nal block of mirror image discrimination

trials without the rotational cue (but see results, below). We intended to

use this nal block of trials for a comparison of response latencies on cue

and uncued trials. Collecting these comparison trials from the nal

phase of testing ensured that continued training could not account for

any attening of the latency function observed in the cued trials. Mon-

keys were intended to complete the uncued block of trials after meeting

the same performance criterion used in Experiment 1.

We rst compared performance on the last 10 sessions of cued mirror

image trials with performance on the uncued mirror image trials

collected during Experiment 1. We also compared these same cued

mirror image trials with a nal block of 10 sessions of uncued mirror

image trials collected as the last part of Experiment 3.

During the initial administration of Experiment 3, a programming

error caused the rotational cue to disappear on the majority of trials

immediately after monkeys touched the sample. The coding error also

affected randomizing of the left and right position of correct choices,

allowing monkeys to complete the task without the aid of the cue. The

programming error was only noticed after monkeys had completed

Experiment 3. We therefore retrained monkeys to use the cue by having

Table 1

Accuracy across rotations for the last 6 sessions of Experiment 2. All 6 monkeys

were signicantly more accurate than expected by chance as determined by

binomial tests.

Orientation (Degrees) of test images

◦

120

◦

Subject A 0.96 0.79 0.73 0.82 0.68

B 0.76 0.85 0.69 0.75 0.72

G 0.92 0.79 0.81 0.76 0.64

Sh 0.92 0.85 0.73 0.65 0.66

Sy 0.86 0.82 0.65 0.70 0.82

V 0.92 0.78 0.69 0.69 0.66

T.C. Hassett et al.

Cognition 228 (2022) 105225

them complete the procedures described in Experiment 2, following

which they were given the correct version of Experiment 3 described

above.

Only 1 monkey reached criterion. Two monkeys who were near

criterion were moved from the cued phase of Experiment 3 to the uncued

phase before reaching criterion because the novel coronavirus was set to

interrupt lab operations for an indenite duration. Testing of the nal

three monkeys was paused and resumed later in the year, after labora-

tory operations resumed. After extensive training without fully reaching

criterion (92, 98, and 104 days) the nal three monkeys were moved

from the cued to uncued phase of testing without having met criterion.

All monkeys completed at least 200 sessions, and therefore, we analyzed

their cued test data from sessions 191–200. All six monkeys reached

criterion on the uncued phase of testing.

3.3. Results and discussion

This compound task, intermixing cued and uncued trials unpredict-

ably was clearly difcult for the monkeys. We performed two analyses

for Experiment 3. First, we compared cued mirror image performance to

uncued mirror image performance collected prior to cue training in

Experiment 1. A strength of this comparison is that it allows us to

compare mental rotation performance prior to and after training that

ostensibly caused them to mentally rotate prior to test images appearing.

A weakness of this comparison is that comparatively better performance

on cued trials might be expected simply because the monkeys had more

experience with mental rotation by Experiment 3. Therefore, we per-

formed a second analysis where we compared performance on cued

mirror image trials and uncued mirror image trials collected after cue

training and test, at the very end of Experiment 3, when monkeys had

the maximum practice. The strength of this comparison is that monkeys

experience with mental rotation was approximately equal at the time the

cued and uncued data were collected. A weakness of this comparison is

that monkeys have, at this point, been trained extensively to rotate

images prior to the appearance of tests. If they continue to execute this

rotation whether or not a rotation cue appears, the difference between

cued and uncued trials might not be evident.

Compared to their performance on the uncued Experiment 1 test, the

rotation X response time function was atter in sessions with the rota-

tional cue (Fig. 3 Left; rmANOVA, rotation X cue interaction: F

(1, 5)

28.930, p = .002,

= 0.302; cued vs uncued: F

(1, 5)

= 4.802, p = .079,

= 0.386; rotation: F

(1, 5)

= 49.748, p = .001,

= 0.728). Monkeys

were also overall more accurate on cued trials relative to Experiment 1

(rmANOVA, cued vs uncued: F

(1, 5)

= 13.927, p = .013,

= 0.569;

rotation: F

(1, 5)

= 39.174, p = .001,

= 0.725; cue X rotation interac-

tion: F

(1, 5)

= 0.115, p = .748,

= 0.004). Although this comparison

provides evidence that is consistent with cued mental rotation studies in

humans, it is important to recognize that the Experiment 1 data were

collected at a time when monkeys were comparatively less experienced

at mental rotation tasks. Thus, it is unclear to what extent the observed

signicant differences in latency and accuracy are due to the rotational

cue or simply experience. One reason to believe it is not solely because of

experience is that the cue caused a comparatively greater reduction in

latency with larger rotations compared to smaller rotations. If monkeys

were simply getting better without using the rotation cue, then we

would expect latencies to improve uniformly across degrees of rotation.

Comparing uncued latency functions between Experiment 1 and

Experiment 3 shows that the slope attened signicantly (rmANOVA,

Experiment 1 vs Experiment 3: F(1, 5) = 1.79, p = .239,

= 0.213;

rotation: F(1, 5) = 55.80, p < .001,

= 0.587; test X rotation Inter-

action: F(1, 5) = 11.90, p = .018,

= 0.218).

In contrast to the results reported above, and in contrast to results

from humans, performance on uncued trials following cue training did

not differ from that on cued trials, although the effect of rotation was

still present (Fig. 3 Right; rmANOVA, rotation X cue interaction: F

(1, 5)

0.105, p = .758,

= 0.002; cued vs uncued: F

(1, 5)

= 0.089, p = .776,

= 0.012; rotation: F

(1, 5)

= 37.240, p = .001,

= 0.529; Cooper, 1976;

Cooper & Shepard, 1973; Suchan et al., 2006). Moreover, the rotational

cue did not signicantly increase accuracy (rmANVOA: cued vs uncued:

(1, 5)

= 0.757, p = .424,

= 0.033; rotation: F

(1, 5)

= 21.431, p = .005,

= 0.699; rotation X cue interaction: F

(1, 5)

= 2.138, p = .203,

0.087). Thus, the rst comparison supports the hypothesis that the

rotational cue causes monkeys to rotate images before test, attening the

rotation X response latency function, but the second comparison does

not.

One explanation for why performance did not differ between cued

and uncued trials in the second comparison is that monkeys continued to

mentally rotate images prior to test on both cued and uncued trials, due

to their extensive training with cued trials just prior to these nal tests.

The focus of Experiment 3 was on whether monkeys would use the

rotational cue to mentally rotate in advance of tests. While the evidence

that monkeys could mentally rotate fully in advance of tests is mixed, we

consistently observed longer latencies with larger rotations, replicated

again here. Regardless of whether latency and accuracy functions at-

tened because of the rotational cue or because of experience, monkeys

still showed a relation between extent of rotation and accuracy and la-

tency, and this pattern is characteristic of mental rotation. Throughout

these three experiments, we consistently observed this pattern.

4. General discussion

We found multiple independent signatures of mental imagery,

-0. 2

0.2

0.4

0.6

0.8

Orientation (degrees) of test images

Latency difference (sec) from 0°

Fig. 3. Left: Response times increased less with rotation on cued (gold) than uncued (blue) trials (Experiment 3 Cued vs Experiment 1). Right: Response times

increased equally with rotation on cued (gold) and uncued (blue) trials (Experiment 3). The difference in latency for each rotation from latency at 0

◦

. (For inter-

pretation of the references to colour in this gure legend, the reader is referred to the web version of this article.)

T.C. Hassett et al.

Cognition 228 (2022) 105225

providing converging evidence for the presence of this cognitive process

in monkeys. Monkeys matched remembered upright shapes to rotated

shapes and showed the hallmark longer response time with greater

rotation. Monkeys transformed mental images as instructed by a cue

with precision better than 30

◦

in Experiment 2 and may have used this

cue to mentally rotate images prior to seeing tests in Experiment 3. Thus,

the mental images monkeys formed included ne perceptual details and

remained true to physics through substantial transformation. The nd-

ings from these three experiments, combined with evidence from studies

of blindsight and metacognition that suggest phenomenal visual expe-

rience in monkeys (Andersen et al., 2014; Cowey & Stoerig, 1995; Moore

et al., 2001), indicate that rhesus monkeys form and manipulate mental

images, as dened by the three criteria we dened.

Mental images could underlie a variety of nonhuman animal be-

haviors. Vervet monkeys make specic anti-predator responses when

they hear particular alarm calls and may visualize the predator indicated

by the call in support of this specicity (Cheney & Seyfarth, 1990). It is

common for animals to form expectations of specic rewards, as shown

by selective satiation experiments, and these expectations could involve

visualizing the expected food (Baxter & Murray, 2002). While it may be

considerably less common for nonhumans to have cause to transform

mental images than is the case for humans, tool-users might gain so-

phistication in their use of tools with the aid of isomorphic trans-

formations of mental images, and navigation might be enhanced by

visualization of mental maps (Hunt, 1996; Tolman, 1948; Tomasello,

Davis-Dasilva, Camak, & Bard, 1987).

Human mental representations often include propositional linguistic

content in addition to, or instead of, recapitulating perceptual processes

(Pylyshyn, 1973; Shepard & Cooper, 1982). Lacking language, nonhu-

mans may be especially dependent on representations that are based in

the processes that give rise to the initial perception of stimuli, rather

than propositional representation. Our ndings suggest an evolutionary

continuity, at least among primates, in visual imagery. Evidence from

other cognitive paradigms suggests that nonverbal animals rely on

quasi-visual, rather than propositional, representations to solve a range

of tasks including transitive inference (Gazes et al., 2017; Gazes, Chee, &

Hampton, 2012), quantity discrimination (Brannon & Merritt, 2011;

Gazes et al., 2017; Lourenco & Longo, 2010), and memory for order

(Bunsey & Eichenbaum, 1996; Templer & Hampton, 2013). Visual im-

agery may be an especially powerful form of representation for non-

humans. The evidence presented here showing that monkeys transform

mental images may begin to transform our image of monkey mentality.

CRediT authorship contribution statement

Thomas C. Hassett: Conceptualization, Methodology, Software,

Formal analysis, Investigation, Data curation, Writing – original draft,

Writing – review & editing, Visualization, Project administration. Vic-

toria K. Lord: Conceptualization, Methodology, Formal analysis,

Investigation, Writing – original draft. Robert R. Hampton: Concep-

tualization, Methodology, Investigation, Resources, Writing – original

draft, Writing – review & editing, Supervision, Project administration,

Funding acquisition.

Acknowledgments

We acknowledge support from the National Science Foundation

(BCS-1632477; BCS-1946767), and the National Institutes of Health

(P51OD011132). Tara Dove-VanWormer assisted with testing monkeys.

The authors declare no conicts of interest.

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.

org/10.1016/j.cognition.2022.105225.

References

Adachi, I., Kuwahata, H., & Fujita, K. (2007). Dogs recall their owner’s face upon hearing

the owner’s voice. Animal Cognition, 10, 17–21.

Andersen, L. M., Basile, B. M., & Hampton, R. R. (2014). Dissociation of visual

localization and visual detection in rhesus monkeys (Macaca mulatta). Animal

Cognition, 17, 681–687.

Basile, B. M., & Hampton, R. R. (2013). Dissociation of active working memory and

passive recognition in rhesus monkeys. Cognition, 126, 391–396.

Basile, B. M., Schroeder, G. R., Brown, E. K., Templer, V. L., & Hampton, R. R. (2014).

Evaluation of seven hypotheses for metamemory performance in rhesus monkeys.

Journal of Experimental Psychology. General, 144, 85–102.

Baxter, M. G., & Murray, E. A. (2002). The amygdala and reward. Nature Reviews

Neuroscience, 3(7), 563.

Ben-Haim, M. S., Dal Monte, O., Fagan, N. A., Dunham, Y., Hassin, R. R., Chang, S. W., &

Santos, L. R. (2021). Disentangling perceptual awareness from nonconscious

processing in rhesus monkeys (Macaca mulatta). Proceedings of the National Academy

of Sciences, 118(15).

Brannon, E. M., & Merritt, D. J. (2011). Evolutionary foundations of the approximate

number system. In Space, time and number in the brain (pp. 207–224). Academic

Press.

auer, J., & Belger, J. (2018). A ball is not a Kong: Odor representation and search

behavior in domestic dogs (Canis familiaris) of different education. Journal of

Comparative Psychology, 132, 189–199.

Bunsey, M., & Eichenbaum, H. (1996). Conservation of hippocampal memory function in

rats and humans. Nature, 379(6562), 255.

Burmann, B., Dehnhardt, G., & Mauck, B. (2005). Visual information processing in the

lion-tailed macaque (Macaca silenus): Mental rotation or rotational invariance?

Brain behave evolut, 65, 168–176.

Cheney, D. L., & Seyfarth, R. M. (1990). How monkeys see the world: Inside the mind of

another species. University of Chicago Press.

Cohen, D., & Kubovy, M. (1993). Mental rotation, mental representation, and at slopes.

Cognitive Psychology, 25(3), 351–382.

Cooper, L. A. (1976). Demonstration of a mental analog of an external rotation.

Perception & Psychophysics, 19, 296–302.

Cooper, L. A., & Shepard, R. N. (1973). “Chronometric studies of the rotation of mental

images” in visual information processing (pp. 75–176). Academic Press.

Cowey, A., & Stoerig, P. (1995). Blindsight in monkeys. Nature, 373, 247–249.

Delius, J. D., & Hollard, V. D. (1995). Orientation invariant pattern recognition by

pigeons (Columba livia) and humans (Homo sapiens). Journal of Comparative

Psychology, 109(3), 278–290. https://doi.org/10.1037/0735-7036.109.3.278

Frick, A., Hansen, M. A., & Newcombe, N. S. (2013). Development of mental rotation in

3-to 5-year-old children. Cognitive Development, 28(4), 386–399.

Ganis, G., Thompson, W. L., & Kosslyn, S. M. (2004). Brain areas underlying visual

mental imagery and visual perception: An fMRI study. Cognitive Brain Res, 20,

226–241.

Gazes, R. P., Chee, N. W., & Hampton, R. R. (2012). Cognitive mechanisms for transitive

inference performance in rhesus monkeys: Measuring the inuence of associative

strength and inferred order. Journal of Experimental Psychology. Animal Behavior

Processes, 38(4), 331.

Gazes, R. P., et al. (2017). Spatial representation of magnitude in gorillas and

orangutans. Cognition, 168, 312–319.

Grifn, D. R. (1976). The question of animal awareness: Evolutionary continuity of mental

experience. Rockefeller Univ. Press.

Hampton, R. R. (2001). Rhesus monkeys know when they remember. Proceedings of the

National Academy of Sciences, 98, 5359–5362.

Hampton, R. R. (2009). Multiple demonstrations of metacognition in nonhumans:

Converging evidence or multiple mechanisms? Comp cogn behave rev, 4, 17–28.

Hollard, V. D., & Delius, J. D. (1982). Rotational invariance in visual pattern recognition

by pigeons and humans. Science, 218, 804–806.

Hopkins, W. D., Fagot, J., & Vauclair, J. (1993). Mirror-image matching and mental

rotation problem solving by baboons (Papio papio): Unilateral input enhances

performance. Journal of Experimental Psychology. General, 122, 61–72.

Hunt, G. R. (1996). Manufacture and use of hook-tools by new Caledonian crows. Nature,

379(6562), 249.

ohler, C., Hoffmann, K. P., Dehnhardt, G., & Mauck, B. (2005). Mental rotation and

rotational invariance in the rhesus monkey (Macaca mulatta). Brain behave evolut,

66, 158–166.

Kosslyn, S. M. (1980). Image and mind. Harvard University Press.

Kosslyn, S. M. (1988). Aspects of a cognitive neuroscience of mental imagery. Science,

240, 1621–1626.

Lohmann, A., Delius, J. D., Hollard, V. D., & Friesel, M. F. (1988). Discrimination of

shape reections and shape orientations by Columba livia. Journal of Comparative

Psychology, 102(1), 3–13. https://doi.org/10.1037/0735-7036.102.1.3

Lourenco, S. F., & Longo, M. R. (2010). General magnitude representation in human

infants. Psychological Science, 21(6), 873–881.

Mauck, B., & Dehnhardt, G. (1997). Mental rotation in a California Sea lion (Zalophus

californianus). The Journal of Experimental Biology, 200, 1309–1316.

Miller, E. K., Erickson, C. A., & Desimone, R. (1996). Neural mechanisms of visual

working memory in prefrontal cortex of the macaque. The Journal of Neuroscience,

16, 5154–5167.

Moore, T., Rodman, H. R., & Gross, C. G. (2001). Direction of motion discrimination after

early lesions of striate cortex (V1) of the macaque monkey. Proceedings of the

National Academy of Sciences, 98, 325–330.

Moulton, S. T., & Kosslyn, S. M. (2009). Imagining predictions: Mental imagery as mental

emulation. Phil Trans R Soc B, 364, 1273–1280.

T.C. Hassett et al.

Cognition 228 (2022) 105225

Neiworth, J. J., & Rilling, M. E. (1987). A method for studying imagery in animals. J Exp

Psychol Anim B, 13, 203–214.

Nekovarova, T., Nedvidek, J., Klement, D., Rokyta, R., & Bures, J. (2013). Mental

transformations of spatial stimuli in humans and in monkeys: Rotation vs.

translocation. Behavioural Brain Research, 240, 182–191.

Newcombe, N. S. (2002). The nativist-empiricist controversy in the context of recent

research on spatial and quantitative development. Psychological Science, 13(5),

395–401.

Parr, L. A., & Heintz, M. (2008). Discrimination of faces and houses by rhesus monkeys:

The role of stimulus expertise and rotation angle. Animal Cognition, 11, 467–474.

https://doi.org/10.1007/s10071-008-0137-4

Pearson, J., & Kosslyn, S. M. (2015). The heterogeneity of mental representation: Ending

the imagery debate. Proceedings of the National Academy of Sciences, 112,

10089–10092.

Premack, D. (1983). The codes of man and beasts. The Behavioral and Brain Sciences, 6,

125–136.

Pylyshyn, Z. W. (1973). What the mind’s eye tells the mind’s brain: A critique of mental

imagery. Psychological Bulletin, 80(1), 1.

Shepard, R. N., & Cooper, L. A. (1982). Mental images and their transformations. The MIT

Press.

Shepard, R. N., & Metzler, J. (1971). Mental rotation of three-dimensional objects.

Science, 171, 701–703.

Suchan, B., Botko, R., Gizewski, E., Forsting, M., & Daum, I. (2006). Neural substrates of

manipulation in visuospatial working memory. Neuroscience, 139, 351–357.

Templer, V. L., & Hampton, R. R. (2013). Cognitive mechanisms of memory for order in

rhesus monkeys (Macaca mulatta). Hippocampus, 23(3), 193–201.

Tolman, E. C. (1948). Cognitive maps in rats and men. Psychological Review, 55(4), 189.

Tomasello, M., Davis-Dasilva, M., Camak, L., & Bard, K. (1987). Observational learning

of tool-use by young chimpanzees. Journal of Human Evolution, 2(2), 175–183.

Vauclair, J., Fagot, J., & Hopkins, W. D. (1993). Rotation of mental images in baboons

when the visual input is directed to the left cerebral hemisphere. Psychological

Science, 4, 99–103.

Zakrzewski, A. C., Johnson, J. M., & Smith, J. D. (2017). In J. Call, G. M. Burghardt,

I. M. Pepperberg, C. T. Snowdon, & T. Zentall (Eds.), “The comparative psychology of

metacognition” in APA handbooks in psychology. APA handbook of comparative

psychology: Perception, learning, and cognition (pp. 703–721). Washington, DC, US:

American Psychological Association.

T.C. Hassett et al.