A

psccosta©

K

1

lefummscs[mafi“i

0d

Behavioural Brain Research 172 (2006) 219–232

Research report

Visual search for moving and stationary items in chimpanzees(Pan troglodytes) and humans (Homo sapiens)

Toyomi Matsuno ∗,1, Masaki TomonagaPrimate Research Institute, Kyoto University, Inuyama, Aichi 484-8506, Japan

Received 28 October 2005; received in revised form 2 May 2006; accepted 4 May 2006Available online 21 June 2006

bstract

Four visual search experiments were conducted using human and chimpanzee subjects to investigate attentional processing of movement, anderceptual organization based on movement of items. In the first experiment, subjects performed visual searches for a moving target amongtationary items, and for a stationary target among moving items. Subjects of both species displayed an advantage in detecting the moving itemompared to the stationary one, suggesting the priority of movement in the attentional processing. A second experiment assessed the effect of theoherent movement of items in the search for a stationary target. Facilitative effects of motion coherence were observed only in the performancef human subjects. In the third and fourth experiments, the effect of coherent movement of the reference frame on the search for moving and

tationary targets was tested. Related target movements significantly influenced the search performance of both species. The results of the second,hird, and fourth experiments suggest that perceptual organization based on coherent movements is partially shared by chimpanzees and humans,nd is more highly developed in humans.

2006 Elsevier B.V. All rights reserved.

t mov

fiboiino

trcpes

eywords: Chimpanzee; Perceptual organization; Search asymmetry; Coheren

. Introduction

Detecting a moving item is highly important for animalsiving in dynamically changing visual environments, and thisxplains why our perceptual mechanisms are so well-developedor that purpose. The visual search paradigm has frequently beensed to investigate visual attention mechanisms for processingotion information [8,16,20,31,32]. For example, search asym-etry for motion was found in studies that tested human visual

earch performance in two display conditions, one of whichonsisted of a moving dot and stationary dots and the otherymmetrically designed with a stationary dot and moving dots37,54]. Results indicated that detecting a stationary item amongoving items is more difficult than detecting a moving item

mong stationary items. Royden and colleagues explained these

ndings according to feature integration theory [49], such thatmotion” is a basic feature in our visual system and “stasis” ists absence.

∗ Corresponding author. Tel.: +81 568 63 0567; fax: +81 568 63 0550.E-mail address: matsuno@pri.kyoto-u.ac.jp (T. Matsuno).

1 JSPS Research Fellow.

veinp

p

166-4328/$ – see front matter © 2006 Elsevier B.V. All rights reserved.oi:10.1016/j.bbr.2006.05.004

ement; Reference frame

When we think about detecting a target in a dynamic visualeld, perceptual organization is also an important considerationecause an item’s motion is often perceived in the global contextf the movement of other items. For example, our visual systems sensitive to movement coherence [56]; coherently movingtems are more easily grouped [24]. Also, as is apparent in phe-omena such as induced or relative motion, an object’s motion isften perceived in relationship to the movements of other items.

This perceptual organization of moving items and its relationo visual search performance has been investigated by severalesearchers [9,23,55]. In studies using human subjects, motionoherence of the search items was varied and effects on searcherformance tested. These studies found that the effect of coher-nt motion was to perceptually group the items as an organizedurface, probably processed in the “preattentive” stage of humanisual perception [48], and that such perceptual grouping influ-nced the search for a moving target. Results of these studiesndicated a strong tendency for perceptual grouping or orga-

izing of moving items in ways that affected later attentionalrocessing of the items.

In the field of comparative cognition, studies on visualerception in non-human animals have been important in under-

2 ural B

smibreh[soaimidp

aiiin

tmhscitm

dzthmst

mempm

2

2

MyEi

ts

rtphthv

ooTI

ir

2

1witsiMrtrawiSv(

bTt

3m

itsom

3

3

st

3

20 T. Matsuno, M. Tomonaga / Behavio

tanding how non-human animals perceive their world, and toanifest the evolutionary foundation of human visual process-

ng [2,46]. Several learning experiments have shown similaritiesetween humans and chimpanzees, our closest evolutionaryelatives, in visual perception and attentional processing. Forxample, chimpanzees showed comparable performance toumans in characteristics of early vision, such as visual acuity30] and color and brightness perception [14]. In a series of visualearch experiments, Tomonaga [46] examined features such asrientation, form, texture pattern, and shape-from-shading cues,nd demonstrated “pop out” and search asymmetry phenomenan chimpanzees, similar to those in humans. Perception of

ovement in visual search, however, has been less frequentlynvestigated in non-human animals, although information aboutetection and attentional processing of moving items wouldrovide highly important ecological and evolutionary validity.

There has also been little investigation into how non-humannimals organize their visual perception of objects. Whereast is natural for humans to perceive the relationship betweentems in a visual field [24], previous reports suggest that thiss not always true for non-primate animals [5,25,52], or evenon-human primates, such as chimpanzees [11,13].

The purpose of this study was two-fold. First, we soughto reveal whether chimpanzees showed search asymmetry for

oving and stationary targets. The question was whether, likeumans, chimpanzees more rapidly detect a moving item than atationary item. In Experiment 1, we used a visual search task toompare the performance of humans and chimpanzees in detect-ng a moving target among stationary distractors, and a stationaryarget among moving distractors. This allowed us to measure the

ost basic visual processing tendencies of chimpanzees.Second, we examined the organization of visually perceived

iscrete moving objects. Previous studies of perceptual organi-ation in chimpanzees used stationary stimuli [12]; we attemptedo test the dynamic aspect of perceptual organization in non-uman primates. In Experiment 2, a stationary item amongoving distractors was presented to the subjects under two

earch conditions in which the uniformity of the movement ofhe distractors was varied.

To further investigate perceptual organization, search perfor-ance for moving and stationary targets as in Experiment 1 was

valuated in Experiment 3 (3a and 3b), with the addition of aoving reference frame. These two experiments examined the

erceptual organization of discrete items in the context of relatedovement.

. General methods

.1. Subjects

Five chimpanzees: Ai (27 years old, female), Akira (28 years old, male),ari (28 years old, female), Pendesa (27 years old, female), and Ayumu (3.5

ears old, male), participated in the experiments. Ai and Akira participated in

xperiments 1, 3a, and 2 in that order. Mari, Pendesa, and Ayumu participated

n Experiment 3b.The subjects were experienced in performing various perceptual-cognitive

asks. At the time of the present study, Ai was experienced in matching-to-ample-tasks, such as symbolic and identity matching tasks [1,27,29], and had

(a2o

rain Research 172 (2006) 219–232

elatively little experience in visual search tasks [43,44]. Akira had been highlyrained in visual search tasks [42,45,47]. However, neither of these subjects hadrior experience of the task used in the present study. Mari and Pendesa alsoad experience in performing discrimination tasks [1,27,41], and had learnedhe visual search for a moving target task [28] in advance of this study. Ayumuad little experience in computer-controlled tasks, but he had also learned theisual search task that used a moving target among stationary discs [28].

The subjects lived with 10 other chimpanzees in an environmentally enrichedutdoor compound and attached indoor residences [33]. They were not deprivedf food at any time during the study. Care and use of the chimpanzees adhered tohe Guide for the Care and Use of Laboratory Primates of the Primate Researchnstitute, Kyoto University, Japan.

In addition to the chimpanzees, human volunteers participated in the exper-ments. They were not informed about the purpose of the experiments. All wereight-handed and reported normal or corrected-to-normal visual acuity.

.2. Apparatus

Chimpanzees were tested in an experimental booth (approximately.8 m × 1.8 m × 2.0 m) with acrylic panels as walls on all four sides. Stimuliere generated on a Pentium-based computer and displayed on 21-in and 22-

n CRT monitors (Totoku CV-213PJ for Ayumu and Mitsubishi TSD-221S forhe other subjects) equipped with capacitive and surface acoustic wave touchcreens. This monitor system served to present the stimuli, and was also thenput device for subject responses with accurate information for touch locations.

onitor resolution was 1024 × 768 pixels with 8-bit color mode. The refreshate was 75 Hz and the display was synchronized with the vertical retrace ofhe monitor. The positions of the moving stimuli were updated on every screenetrace to give the impression of smooth motion. Subjects observed the monitort a viewing distance of about 40 cm without head restraint. The viewing distanceas roughly restricted by an acrylic panel, which was attached between the mon-

tor and subjects to prevent the destruction of the monitor by the chimpanzees.timulus luminance was measured using a colorimeter (Topcon, BM-7). A uni-ersal feeder (Biomedica, BUF-310) delivered small pieces of a food rewardapples or raisins) into a food tray below the monitor.

Human subjects were tested with the identical apparatus in the experimentalooth. The only exception was that they were not rewarded with pieces of food.hey were required to observe the monitor from a distance of about 40 cm and

o respond with a finger touch as did chimpanzees.

. Experiment 1: asymmetry in visual search foroving and stationary targets

Experiment 1 tested visual search for moving and stationarytems. Targets and distractors were designed symmetrically inwo search display conditions (Fig. 1). In one condition, subjectsearched for a moving target among stationary items, and in thether, they were required to detect a stationary target amongoving distractors.

.1. Methods

.1.1. SubjectsTwo chimpanzees (Ai and Akira), and five undergraduate

tudents (three males and two females) ranging in age from 18o 22 years (mean = 19.2 years), participated in Experiment 1.

.1.2. StimuliThe total screen area subtended 392 mm × 292 mm

52.2◦ × 40.1◦ of visual angle at a viewing distance of 40 cm),nd the maximum display area (with 12-item display) was30 mm × 172 mm (32.1◦ × 24.3◦), excluding the lower partf the screen where a warning stimulus, which showed the

T. Matsuno, M. Tomonaga / Behavioural B

iDvu

sbdoci(f

mtgam

mtfissatws

3

atsomwttdttttsTatr

tpfaccu

w2bstimulus detection, movement preparation, and movement exe-cution. This also served maintain a high reward rate and sustainthe motivation of chimpanzees during test sessions. The firsteight trials (two trials for each display size) of a test session

1 In the initial phase of pilot training, one of the chimpanzee subjects, Ai, wastrained in an odd-item search paradigm [32], in which she was required to detectthe single moving or stationary target among five distractors without the samplepresented in advance of the search display. After 10 sessions of 32 trials foreach display condition of the training, Ai did not show any evidence of learningand her motivation decreased to stop the task because of low reward rate. We

Fig. 1. Examples of search displays used in Experiment 1.

nitiation of each trial, and a sample stimulus were presented.isplays were comprised of 1, 4, 8, or 12 items (display-sizeariable) including the target, and were presented continuouslyntil terminated by the subject’s response.

The stimuli were black discs (approximately 15 cd/m2)ubtending about 12 mm × 12 mm (1.7◦ × 1.7◦) against a grayackground (approximately 30 cd/m2). They were randomlyistributed on an imaginary 4 × 3 square matrix with cell sizef 57 × 57 mm (8.2◦ × 8.2◦), subject to the constraint that eachell contained no more than one item. The initial position of antem in a cell was also randomly set within a 38 mm × 38 mm5.5◦ × 5.5◦) area centered in the cell, so that items neverormed orderly vertical or horizontal lines.

In the moving target condition, a display consisted of aoving disc (target) and stationary discs. In the stationary

arget condition, a display consisted of a stationary disc (tar-

et) and moving discs. The stimuli oscillated horizontally atvelocity of 57 mm/s (8.2◦/s). All moving items in a displayoved at the same velocity and reversed their direction of

tscs

rain Research 172 (2006) 219–232 221

otion simultaneously after every migration of 12 mm (1.8◦);hus, the stimuli moved in phase. Each human subject con-rmed that the moving items produced the impression of amooth continuous motion, and that they did not leave a per-istently visible trail. Each item never passed beyond the cell,nd the minimum separation from an adjacent item was main-ained at more than 28 mm (center-to-center). The laboratoryas dimly illuminated to prevent reflections on the computer

creen.

.1.3. ProcedureA delayed matching-to-sample (DMTS) task with multiple

lternatives [43] was used.1 Each trial was initiated by the simul-aneous presentation of a warning stimulus (an empty blackquare subtending 30 mm × 30 mm) located at the bottom rightf the screen, and a sample stimulus, which had the same move-ent state as the target, at the bottom center of the screen. Thearning and sample stimuli disappeared after they were sequen-

ially touched. After 500 ms from when the sample was touched,he search display was presented. The locations of the target andistractors were randomly selected from the 12 cells in eachrial. A touch response to an item was defined as a detectedouch within an invisible rectangle (28 mm × 28 mm) aroundhe center of the item. The area moved with the movement ofhe item. When subjects correctly touched the target, a chimeounded, and for the chimpanzees, a food reward was delivered.he choice of incorrect items was followed by a buzzer soundnd a 3-s timeout. The time interval between the presentation ofhe search display and the touch of the item was recorded as theesponse time.

Prior to the test sessions, the chimpanzees were trained onhe search task for moving and stationary targets in the dis-lay size 6 condition. A session consisted of 64 trials (32 trialsor each target condition). The criterion for learning was sets >90% accuracy in three consecutive sessions for each targetondition. When performance reached the criterion in a targetondition, intensive training in the other condition was contin-ed until reaching its criterion.

During the test phase, a session consisted of 104 trials inhich the display size varied among 1, 4, 8, and 12 items, with6 trials for each display size. Display size 1 was used to collectaseline chronometric information reflecting the processes of

hen introduced the delayed matching-to-sample (DMTS) task, with which theubject had been familiar thorough long-term continuous training. The otherhimpanzee subject, Akira, started his training in the DMTS task with displayize 6, and succeeded in learning it.

2 ural B

wqigdnd

Auo

s1Asifit

fs(utprt

3

3

mtsmc(swttmMc

itta(ptrrd

cbaswa

3.2.2. Test phase: response timeThe response time data for display sizes 4–12 in the test

phase are presented in Fig. 2. Searching for a stationary item

22 T. Matsuno, M. Tomonaga / Behavio

ere treated as practice trials and were excluded from subse-uent analyses. The other 24 trials of each display size werentermixed randomly during the session. The position of the tar-et was counterbalanced across the 24 trials. In a session, theisplay condition was fixed, and the two conditions were alter-ately presented. Two consecutive sessions (one session for eachisplay condition) were counted as a test block.

Each chimpanzee subject participated in eight test blocks.ccuracy and response time data from the last six blocks weresed for analyses. Ai began her test sessions with the conditionf a moving target, and Akira began with a stationary target.

Each human participant participated in a single test block (oneession for each display condition) with two practice sessions of0 trials (one session for each display condition; display size 6).practice session was given to the subject just prior to the test

ession of the display condition. Human subjects were verballynstructed to correctly and quickly detect the target. Two of theve subjects began their test with a moving target, and the other

hree began with a stationary target.The number of correct responses and median response times

or correct trials were collected for each display size in a ses-ion. During test sessions, averaged accuracy and mean medianMMdn) response time of display sizes 4–12 were analyzedsing two-way analysis of variance (ANOVA; display condi-ion × display size); repeated measures were blocks for chim-anzees and subjects for humans. The search slope of theesponse times in the test sessions was also analyzed using two-ailed t-tests.

.2. Results

.2.1. Training phaseIn the display size 6 training phase, Akira took longer to

eet the criterion in a stationary target condition than a movingarget condition (8 sessions for a moving target condition; 23essions for a stationary target condition). He performed muchore accurately in a moving target condition (mean percentage

orrect = 77.3, S.E. = 8.05) than in a stationary target conditionmean percentage correct = 9.38, S.E. = 2.83) in the first eightessions, t(7) = 6.34, p < 0.01. This suggests that moving targetsere more salient among stationary distractors than stationary

argets among moving distractors for Akira. Correct responseimes were not significantly different between conditions (a

oving target, MMdn = 1167 ms, S.E. = 97; a stationary target,Mdn = 1649 ms, S.E. = 281), partly because of large variance

aused by very few correct trials in a stationary target condition.In contrast, Ai reached the learning criterion more rapidly

n a stationary target condition (29 sessions for a movingarget condition; 21 sessions for a stationary target condi-ion). However, Ai was slower to correctly detect a station-ry target (MMdn = 1181 ms, S.E. = 22) than a moving targetMMdn = 939 ms, S.E. = 27) in the first 21 sessions, t(20) = 6.82,< 0.01, despite no significant difference in accuracy between

he two conditions (stationary target, mean percentage cor-ect = 82.1, S.E. = 2.9; moving target, mean percentage cor-ect = 78.4, S.E. = 1.9), t(20) = 1.37. These results suggest thatetecting a moving target was easier, but that Ai’s selection pro-

Frst

rain Research 172 (2006) 219–232

ess was biased to favor stationary items. This may be partlyecause she had previously worked only on tasks with station-ry stimuli (such as matching tasks using Arabic numerals withtationary dots, color patches, and geometric figures), and thisas the first time she was to select moving items as correct

nswers.

ig. 2. Mean median response times in Experiment 1. Each graph presents theesponse time × display size functions for the two display conditions. Filledquares represent the moving target conditions; open circles represent the sta-ionary target conditions; error bars are ±1 S.E.

ural B

amhcdffbpoa1ta8tRcAFf

s1mftw(m

3

cstFtecap

TMa

D

M

S

1pmaar1i

pf

3

saaodapatd

acarscftilocT

T. Matsuno, M. Tomonaga / Behavio

mong moving items was markedly more difficult than finding aoving target among stationary items for both chimpanzees and

umans. There was, however, a remarkable discrepancy betweenhimpanzees and humans. Response time differences betweenisplay conditions increased as a function of the display sizeor chimpanzees, but stayed relatively constant for human per-ormance. The main effect of display condition was significantoth in chimpanzees, F(1, 5) = 48.88 for Ai and 88.1 for Akira,< 0.01, and humans, F(1, 4) = 193.42, p < 0.01. The main effectf display size was significant for Ai, F(2, 10) = 6.20, p < 0.01,nd humans, F(2, 8) = 4.90, p < 0.05, but not for Akira, F(2,0) = 0.43. There was a significant interaction of display condi-ion and display size for the chimpanzees, F(2, 10) = 5.86 for Aind 10.90 for Akira, p < 0.01, but not for the human subjects, F(2,) = 3.10, indicating that the simple main effect of display condi-ion was significantly modified by display size for chimpanzees.esponse times were not significantly different between displayonditions in display size 4, F(1, 15) = 3.48 for Ai and 1.45 forkira, but were significantly different in display sizes 8 and 12,(1, 15) = 14.16 and 44.11 for Ai; F(1, 15) = 34.92 and 61.86

or Akira, p < 0.01.Search slopes in the detection of a stationary target (mean

earch rate = 70.11 ms/item, S.E. = 22.07, for Ai, 42.80 ms/item,7.01, for Akira) were higher than those in the detection of aoving target (mean search rate = −0.85 ms/item, S.E. = 2.47

or Ai, −72.09 ms/item, 24.70 for Akira) in chimpanzees,(5) = 3.03 for Ai, p < 0.05, and 4.14 for Akira, p < 0.01, but theyere not significantly different between conditions in humans

stationary target, mean search rate = 7.78 ms/item, S.E. = 3.24;oving target, 0.64 ms/item, 0.84), t(4) = 1.91, p > 0.10.

.2.3. Test phase: accuracyChimpanzees performed much more accurately than due to

hance in the test sessions, and the mean percentage correcthowed similar patterns to response time (Table 1). For Ai,here were no significant main effects of display condition,(1, 5) = 0.13, or display size, F(2, 10) = 1.03, but the interac-

ion was significant, F(2, 10) = 7.20, p < 0.05. The simple main

ffects analysis revealed that performance in a stationary targetondition was significantly more accurate than performance inmoving target condition for display size 4, F(1, 15) = 5.84,< 0.05, and that the tendency was reversed for display size

able 1ean percentage of correct responses with the standard error for each subject

nd each condition in Experiment 1

isplay size Ai Akira

% Correct S.E. % Correct S.E.

otion4 78.5 4.5 86.1 2.38 84.0 5.0 91.0 1.312 95.8 2.2 88.9 3.5

tasis4 96.5 1.7 72.2 3.78 92.4 1.3 89.6 3.212 73.6 10.2 87.5 2.8

ipdr

avmptcddwrtci

rain Research 172 (2006) 219–232 223

2, F(1, 15) = 8.81, p < 0.01. For Akira, the main effect of dis-lay condition was significant, F(1, 5) = 7.50, p < 0.05, reflectingore accurate performance in a moving target condition than instationary target condition. The main effect of display size waslso significant, F(2, 10) = 16.32, p < 0.01, indicating less accu-ate performance in display size 4 than in display sizes 8 and2 (post-hoc comparisons using Ryan’s method, p < 0.05). Thenteraction was not significant, F(2, 10) = 2.78.

Human subjects exhibited almost perfect performance (meanercentage correct = 99.3), so their accuracy was not analyzedor statistical significance because of possible ceiling effects.

.3. Discussion

Both chimpanzees and humans demonstrated asymmetricalearch performance; it was more difficult to find a stationary itemmong moving distractors than a moving target among station-ry stimuli. These results are consistent with previous studiesn human search performance that showed a clear advantage ofetecting a moving target [37,54], and imply that chimpanzeesnd humans share the same mechanism of visual attention thatrocesses motion. These results are compatible with evolution-ry theories because the ability to rapidly shift visual attentiono moving items may have a major survival value for visuallyependent species like chimpanzees and humans.

Search asymmetry itself was observed both in chimpanzeesnd humans, but the degree of asymmetry was much greater inhimpanzees. In humans, the rates to detect moving and station-ry targets were almost the same, with little or no increment inesponse times as a function of display size. In fact, the searchlopes showed an efficient level (<10 ms/item) [57] under bothonditions in humans. The search for a moving target was per-ormed as efficiently by the chimpanzees as the humans, andhere was no deterioration in performance with an incrementn the number of distractors. Performance was facilitated witharger display sizes, suggesting that a moving target “poppedut” among dense and uniform background items in the per-eptual analysis of chimpanzees, as it did for humans [2,42].he search for a stationary item, however, was not as efficient

n chimpanzees, and the difference in performance among dis-lay conditions was much greater with larger display sizes. Thisifference in the performance of chimpanzees and humans mayeflect species differences in visual processing.

Even in humans, the search for a stationary target is notlways performed efficiently. Royden et al. [37] investigatedisual search performance for a stationary item under threeotion conditions. In the Uniform motion condition, all dis-

lay items moved in phase as in the present experiment; inhe other two conditions, the Random and Brownian motiononditions, the items moved out of phase to randomly selectedirections. They found that response times in the Uniform con-ition were only slightly or not at all affected by display size,hich is consistent with the results of the present study, whereas

esponse times in the other two conditions increased as a func-ion of increased display size, consistent with the results for ourhimpanzee subjects. They suggested that perceptually group-ng uniformly moving distractors, or an induced motion effect

2 ural B

cttiab

psw

4a

dsIuvm

4

4

Esfe

4

idso1mtItpoth

4

sfisuc

aa

ssc

eTs

AS

4

4

racisuhcAFarsb

asrtpstSaSt

4

tfio1

(

4

24 T. Matsuno, M. Tomonaga / Behavio

aused by their uniformity, would make it easier to detect a sta-ionary target. The discrepancy in search performance betweenhe two species we tested could thus be explained by differencesn perceiving uniformly moving distractors. Humans could takedvantage of distractor uniformity to efficiently detect a target,ut this would be more difficult for chimpanzees.

The effect of uniform or coherent motion on visual searcherformance has been investigated in humans [9,23,55], but nouch studies have been conducted in chimpanzees. Therefore,e addressed this issue in Experiment 2.

. Experiment 2: visual search for a stationary targetmong uniformly or randomly moving distractors

We conducted a search for a stationary target among movingistractors task under two conditions. One condition was theame as in Experiment 1, where all distractors moved in phase.n the other condition, the distractors moved out of phase. If theniform motion of the distractors facilitates search performance,isual search in the former condition, i.e., when the distractorsoved in phase, would be easier.

.1. Methods

.1.1. SubjectsThe same chimpanzees from Experiment 1 participated in

xperiment 2. Newly recruited four graduate and undergraduatetudent subjects (one male and three females) ranging in agerom 18 to 24 years (mean = 20.8 years) also participated in thexperiment.

.1.2. StimuliThe stimuli used in Experiment 2 were the same as in Exper-

ment 1 except as reported here. In the Uniform condition, theistractors oscillated horizontally in phase. Movements of thetimuli were the same as those in Experiment 1 except for thescillation amplitude (18 mm, 2.6◦) and display size (1, 3, 7, and1). The oscillation amplitude was enlarged in order to allowore variations of oscillation phase in Random condition and

o make the difference between two conditions more apparent.n the Random condition, half of the distractors moved horizon-ally and the others moved vertically. In addition, the oscillationhases of all distractors varied randomly, so that the movementf distractors appeared disorganized. The amplitudes of the ver-ical and horizontal oscillations were the same as those of theorizontal oscillations in the Uniform condition.

.1.3. ProcedureA test session consisted of 104 trials (26 trials for each display

ize), in which the display condition (Uniform or Random) wasxed. Each chimpanzee participated in eight test blocks (eightessions for each display condition), and the last six blocks weresed for analyses. Ai started her test sessions with the Random

ondition and Akira started his with the Uniform condition.

The test sessions were presented to the chimpanzees withoutny additional training or practice. Experiment 2 was conductedfter Experiment 3a to avoid the additional sessions biased for a

trt

rain Research 172 (2006) 219–232

tationary target search prior to Experiment 3a, in which theearch performances for moving and stationary targets wereompared as in Experiment 1.

Each human subject performed one test block (one session forach display condition) with two practice sessions of 10 trials.he order of the tested conditions was counterbalanced betweenubjects.

Accuracy and response time were analyzed using two-wayNOVA (display condition × display size) as in Experiment 1.earch slopes were also analyzed using one-tailed t-tests.

.2. Results

.2.1. Response timeBoth chimpanzees showed monotonically increasing

esponse time as a function of display size in both the Uniformnd Random conditions (Fig. 3). In the performance of bothhimpanzees, only the main effect of display size was signif-cant, F(2, 10) = 57.31 for Ai and 10.77 for Akira, p < 0.01,uggesting that searching for a stationary target was inefficientnder both display conditions for chimpanzees. In contrast,uman response times in the Uniform condition were relativelyonstant compared to those in the Random condition. Two-wayNOVA revealed significant main effects for display condition,(1, 3) = 12.86, p < 0.05, display size, F(2, 6) = 10.41, p < 0.05,nd their interaction, F(2, 6) = 6.03, p < 0.05. Post-hoc analysesevealed simple main effects of the display condition for displayizes 7 and 11, F(1, 9) = 12.37 and 16.38, respectively, p < 0.01,ut not for display size 3, F(1, 9) = 0.04.

Analyses of search slopes also revealed a discrep-ncy between chimpanzees and humans. Humans showedteeper search slopes in the Random condition (mean searchate = 12.22 ms/item, S.E. = 1.17) than in the Uniform condi-ion (mean search rate = 4.20 ms/item, S.E. = 2.59), t(3) = 2.78,< 0.05, as reported in a previous study [37]. In contrast, search

lopes of chimpanzees were not significantly different betweenhe Uniform condition (mean search rate = 40.36 ms/item,.E. = 3.42, for Ai and 51.07 ms/item, 16.21, for Akira)nd the Random condition (mean search rate = 55.59 ms/item,.E. = 9.85, for Ai and 48.56 ms/item, 12.68, for Akira),

(5) = 1.64 for Ai and 0.26 for Akira, p > 0.10.

.2.2. AccuracyChimpanzees maintained a very high level of performance

hroughout the sessions (Table 2), partly because the target wasxed to a stationary item during testing. Only the main effectf the display size in Akira’s performance was significant, F(2,0) = 7.60, p < 0.01, consistent with the results for response time.

Human subjects exhibited almost perfect performanceM = 98.8). The data were not analyzed further.

.3. Discussion

The results of Experiment 2 supported the hypothesis thathe different tendencies in human and chimpanzee performanceevealed in Experiment 1 were partly due to the uniformity ofhe moving distractors. The results of human performance were

T. Matsuno, M. Tomonaga / Behavioural Brain Research 172 (2006) 219–232 225

Fig. 3. Mean median response times in Experiment 2. Each graph presents thercr

crtcgItipst

Table 2Mean percentage of correct responses with the standard error for each subjectand each condition in Experiment 2

Display size Ai Akira

% Correct S.E. % Correct S.E.

Uniform3 98.6 1.4 93.1 1.87 97.2 1.4 97.9 1.411 93.8 3.8 92.4 2.0

Random3 94.4 4.8 91.7 1.1

gtf

pSKteoeifiwme

onecrie

pcTea

5m

tts

esponse time × display size functions for the two display conditions. Openircles represent the conditions of uniformly moving distractors; filled lozengesepresent the conditions of randomly moving distractors; error bars are ±1 S.E.

onsistent with the study by Royden et al. [37], showing shorteresponse times and more efficient search rates in the Uniformhan in the Random condition. This advantage for the Uniformondition suggests that human subjects perceptually organize aroup of distractors depending on the uniformity of their motion.n contrast, the chimpanzees did not demonstrate such an advan-age; performances in both conditions were the same, with a clear

ncrement in response time as a function of display size, com-arable to human performance in the Random condition. Thisuggests that chimpanzees did not have the significant advan-age of perceptual grouping by uniform motion, nor could they

tiif

7 97.2 1.8 97.9 1.411 95.1 2.7 93.8 0.9

lobally process motion coherence in this task, although the rela-ively large drop seen in response time for the Uniform conditionor Ai at display size 11 implied a similar but weaker tendency.

Previous studies using stationary stimuli also report restrictederceptual organization in chimpanzees compared to humans.tudies by Fagot and Tomonaga [13] on the perception of theanizsa illusory figure found that chimpanzees were more sensi-

ive than humans to the separation between four Pacman-shapedlements within a display, and that the illusory effect disappearednly for the chimpanzees when the distance between the visuallements was enlarged. Fagot and Tomonaga [12] also stud-ed global and local processing in chimpanzees using geometricgures comprised of smaller geometric elements and found thathen the density of the elements was sparse, chimpanzee perfor-ance shifted to local precedence, while humans consistently

xhibited global precedence.Chimpanzees can, however, perceptually organize visual

bjects in a display; when the separation between elements wasot so large, chimpanzees could perceptually organize the visuallements and perceive the illusory square [13]. This was also thease for the advantage of global processing [17,18]. Given theseesults, it is possible that chimpanzees differ from humans onlyn the degree with which they can perceptually organize visuallements.

We reasoned that, if the visual stimuli are easier to organizeerceptually, the chimpanzees should take advantage of the per-eptual grouping of moving items in their search performance.hus, in the next experiments, we further investigated the influ-nce of perceptual organization on the visual search for movingnd stationary targets.

. Experiment 3a: asymmetry reversal with frameotion

To further investigate the ability of chimpanzees to percep-ually organize coherently moving items in the visual field, weested the effect of movement of the reference frame on theearch for moving and stationary targets. In addition to using

ests similar to those in Experiment 1 (visual search for a mov-ng or a stationary target with a stationary reference frame), wentroduced two new display conditions; these required searchingor a moving or a stationary target among distractors within a

2 ural Brain Research 172 (2006) 219–232

sdtso

ctcdIsffcdhe(i

5

5

ps2Ot

5

usccaBc

o(iMm1

fwef

5

As

FEa

rtfswt

sfca

s

26 T. Matsuno, M. Tomonaga / Behavio

ynchronously moving reference frame. The mutual coherenceue to proximity or inclusiveness between the moving items andhe reference frame used in this experiment was expected to betronger than the coherence that was present in the relationshipf the discrete discs in Experiment 2.

Strong coherence of the reference frame with the target itemsould influence the perception of the relative movements of thearget items. Human subjects reported that they perceived theoherent movement of discs with the reference frame as if theiscs were settled on the surface of the moving reference frame.f chimpanzees also perceived the relative state of the target in aimilar fashion, the influence of the movement of the referencerame on detecting moving and stationary targets would be dif-erent because the interaction of the reference frame and targetould change the relative position of the target to its oppositeirection. If chimpanzee perception was dissimilar from that ofumans, the movement of the reference frame would not influ-nce search performance or the influence would be constantmerely disturbing), rather than provide relative and positionalnformation.

.1. Methods

.1.1. SubjectsThe same chimpanzees from Experiments 1 and 2 partici-

ated in Experiment 3a. Six graduate and undergraduate studentubjects (one male and five females) ranging in age from 19 to6 years (mean = 22.8 years) also participated in Experiment 3.ne of these had previously participated in Experiment 1, and

he other five were newly recruited.

.1.2. StimuliIn Experiment 3a, a reference frame was added to the display

sed in Experiment 1 (Fig. 4). The frame appeared as a grayquare area (approximately 30 cd/m2) with black lines (the sameolor as the background) of 2-mm (0.3◦) width on the borders ofells of 55 mm × 55 mm (7.9◦ × 7.9◦). The frame was presentedgainst an intense black background (approximately 0 cd/m2).lack discs (approximately 15 cd/m2) were presented on grayell areas in the same manner as in Experiment 1.

The movement of the stimuli and frame was a horizontalscillation at a velocity of 57 mm/s (8.2◦/s) and a swing of 12 mm1.7◦). All moving items, including the frame, in a display movedn phase. No item passed over the cell confined by black lines.

inimum separation from the adjacent item was maintained atore than 28 mm (center-to-center). Display size varied among

, 4, 8, and 12 items.Four display conditions (moving frame or stationary

rame × moving target or stationary target) were tested. Whene focused on the relativity of target motion to the refer-

nce frame, the moving frame–moving target and stationaryrame–stationary target conditions were equivalent.

.1.3. ProcedureAs in Experiments 1 and 2, a DMTS procedure was employed.

sample stimulus was presented on the center of the gray squareubtending 42 mm × 42 mm (6.0◦ × 6.0◦), which served as the

tftw

ig. 4. Examples of search displays in the moving reference conditions used inxperiment 2. Moving discs and the reference frame are moving in phase. Therrows depict the oscillation of each element.

eference frame of the sample stimulus at the bottom center ofhe screen. The square moved in the same way as the referencerame. The reference frame and sample stimulus were presentedimultaneously. The sample stimulus and square disappearedhen touched. After a 500-ms interval, a test target and distrac-

ors were presented in the reference frame.A session consisted of 104 trials (26 trials for each display

ize) for each display condition. The first eight trials (two trialsor each display size) were treated as practice trials. Four displayonditions were tested in four consecutive sessions (one block),nd the order within each block was randomly determined.

Each chimpanzee was presented with eight test blocks (eightessions for each display condition). Accuracy and response

imes from the last six blocks (six sessions for each) were usedor analyses. The Experiment 3a test sessions were presentedo the chimpanzees just after the Experiment 1 test sessions,ith no additional training sessions. Each human participant

ural B

wtdcmttA

5

5

osafa

gpcfs

1gcwesf

h5ctyf1dam

t

Ffr

T. Matsuno, M. Tomonaga / Behavio

as presented with a test block (1 session for each) and 4 prac-ice sessions of 10 trials (1 session for each display condition;isplay size 6) immediately prior to the test sessions of theorresponding display condition. Averaged accuracy and meanedian response time for display sizes 4–12 were analyzed using

hree-way ANOVA (reference frame condition × target condi-ion × display size). Search slopes were analyzed using two-wayNOVA (reference frame condition × target condition).

.2. Results

.2.1. Response timeThe response time for display sizes 4–12 revealed the effects

f the reference frame movement (Fig. 5). Akira and the humanubjects exhibited a reversal of their ease in detecting the movingnd stationary targets with the addition of a moving referencerame. On the other hand, Ai did not display that tendency,lthough the degree of search asymmetry was slightly smaller.

There were significant effects of frame condition and tar-et condition, F(1, 5) = 9.59 and 10.02, p < 0.05, for Akira’s

erformance. The two-way interactions were also signifi-ant, F(1, 5) = 215.41 for frame × target; F(2, 10) = 11.50or frame × display size; F(2, 10) = 44.46 for target × displayize, p < 0.01, while the three-way interaction was not, F(2,

c5a1

ig. 5. Mean median response times in Experiment 3a. The three graphs on the leftrame conditions, and the three graphs on the right present the moving reference frameepresent the stationary target conditions; error bars are ±1 S.E.

rain Research 172 (2006) 219–232 227

0) = 0.52. Post-hoc analyses revealed simple main effects of tar-et condition in both the stationary and moving reference frameonditions, F(1, 10) = 69.45 and 5.98 respectively, p < 0.01,hich indicates that detecting a moving target was significantly

asier than detecting a stationary one when the display includedtationary frames, but that the inverse was easier with movingrames.

Similar effects of frame motion were also observed inumans. The main effect of reference frame condition, F(1,) = 11.13, p < 0.05, the interaction of reference frame and targetonditions, F(1, 5) = 24.67, p < 0.01, and the three-way interac-ion were all significant, F(2, 10) = 9.57, p < 0.01. Post-hoc anal-ses revealed simple-simple main effects of the target conditionor reference frame condition and display size, F(1, 30) = 6.41,0.75, 12.09, 12.89, 29.55, and 38.47 for stationary frame-isplay sizes 4, 8, and 12, and moving frame-display sizes 4, 8,nd 12, respectively, p < 0.05, consistent with the search asym-etry reversal in Akira’s performance.Ai did not demonstrate reversed search asymmetry. The

hree-way ANOVA revealed main effects for reference frame

ondition, F(1, 5) = 183.76, p < 0.01, target condition, F(1,) = 178.07, p < 0.01, and display size, F(2, 10) = 35.27, p < 0.01,nd the interaction of target condition and display size, F(2,0) = 23.28, p < 0.01. The other interactions, including refer-

present the response time × display size functions for the stationary referenceconditions. Filled squares represent the moving target conditions; open circles

2 ural Brain Research 172 (2006) 219–232

e5efoad8

sameordtoseS1fmadn1pfrsetsew(sE

5

(i

Fefgea2aeww

Fig. 6. Mean percentage of correct responses for each subject and each conditionin Experiment 3a. The leftmost six bars represent the stationary reference frameconditions, and the others represent the moving reference frame conditions. Thelrf

draWweiat

5

AWii

28 T. Matsuno, M. Tomonaga / Behavio

nce frame and target conditions, were not significant, F(1,) = 4.74 for reference frame × target, F(2, 10) = 0.92 for ref-rence frame × display size; F(2, 10) = 0.96 for referencerame × target × display size. The simple main effects analysisf target condition revealed that Ai took much longer to detectstationary target than a moving target in all display size con-itions, F(1, 15) = 18.50, 72.91, and 175.78 for display sizes 4,, and 12, respectively, p < 0.01.

Search slopes did not show a common tendency amongubjects. Ai had steeper search slopes in the detection of

stationary target (stationary reference frame condition,ean search rate = 46.42 ms/item, S.E. = 4.80; moving refer-

nce frame condition, 41.90 ms/item, 4.37) than in the detectionf a moving target (stationary frame condition, mean searchate = 3.42 ms/item, S.E. = 1.89; moving reference frame con-ition, −2.07 ms/item, 5.31) irrespective of the movement ofhe reference frame, showing only a significant main effectf target condition, F(1, 5) = 70.58, p < 0.01. Akira had higherearch slopes in the stationary target condition (stationary ref-rence frame condition, mean search rate = 142.09 ms/item,.E. = 17.94; moving reference frame condition, 41.53 ms/item,2.75) than in the moving target condition (stationary referencerame condition, mean search rate = −0.52 ms/item, S.E. = 4.35;oving reference frame condition, −113.31 ms/item, 19.22),

nd higher search slopes in the stationary reference frame con-ition than in moving reference frame condition, showing sig-ificant main effects for both conditions, F(1, 5) = 41.39 and46.27 for target and reference frame conditions, respectively,< 0.01. The search slopes of humans showed a significant dif-

erence between the detection of a moving target (mean searchate = 2.09 ms/item, S.E. = 1.03) and a stationary target (meanearch rate = 10.16 ms/item, S.E. = 5.29) only in the moving ref-rence frame condition, showing a significant interaction ofhe two conditions, F(1, 5) = 13.861, p < 0.05, and significantimple main effects of target condition for the moving refer-nce frame condition, F(1, 10) = 10.01, p < 0.05. The differenceas not significant in the stationary reference frame condition

moving target, mean search rate = −2.46 ms/item, S.E. = 1.68;tationary target, 6.66 ms/item, 2.99), F(1, 10) = 1.32, as inxperiment 1.

.2.2. AccuracyHuman subjects exhibited almost perfect performance

M = 98.7). Chimpanzee performance varied, however, accord-ng to the display conditions (Fig. 6).

Akira’s accuracy showed a main effect of display size,(2, 10) = 7.20, p < 0.05, and significant interactions for ref-rence frame × target, F(1, 5) = 153.62, p < 0.01, referencerame × display size, F(2, 10) = 12.52, p < 0.01, and tar-et × display size, F(2, 10) = 10.74, p < 0.01. The simple mainffects of target condition were significant in both the stationarynd moving reference frame conditions, F(1, 10) = 28.95 and1.56, respectively, p < 0.01, indicating more errors detecting

stationary target than a moving target with a stationary ref-

rence frame, and fewer errors in detecting a stationary targetith a moving reference frame. These results were consistentith those of response time.

tdaw

eftmost three of the six bars represent the moving target conditions, and theight three of the six bars represent the stationary target conditions. Each bar isor a different display size (DS) condition. Error bars are ±1 S.E.

Ai’s accuracy showed a main effect of reference frame con-ition, F(1, 5) = 32.11, p < 0.01, and significant interactions foreference frame and target conditions, F(1, 5) = 9.14, p < 0.05,nd target condition and display size, F(2, 10) = 6.47, p < 0.05.

hile there were no simple main effects of the target conditionith a stationary reference frame, F(1, 10) = 0.30, there were

ffects of a moving reference frame, F(1, 10) = 10.24, p < 0.01,ndicating greater accuracy in detecting a stationary target thanmoving target with a moving reference frame, consistent with

he accuracy of Akira’s performance.

.3. Discussion

The search performance of humans and the chimpanzee,kira, showed the same search asymmetry reversal tendency.ith a moving reference frame, Akira displayed obvious facil-

tation in searching for a stationary target and difficulty search-ng for a moving target. These results suggest the possibility

hat chimpanzees can perceive relativity of motion in a searchisplay. In contrast to human performance, which showed a rel-tively inefficient search rate in the detection of a moving targetith a moving reference frame, Akira was quicker to respond,

ural B

rltawaimw

ewsh

6m

cpocmssdts

6

6

itadftdfhittt

6

it

6

dp

st

acwowtcn

6

TimwwtFtMFrtMari, and Pendesa, respectively, p < 0.01, showing an advantageof a moving reference frame in the search for a stationary target.The simple main effects of reference frame in the moving targetcondition were significant only for Mari, F(1, 8) = 8.34, p < 0.05.

Fig. 7. Mean percentage of correct responses for each subject and each conditionin Experiment 3b. The six bars on the left represent the stationary reference frameconditions, and the others represent the moving reference frame conditions. The

T. Matsuno, M. Tomonaga / Behavio

esulting in a negative search slope, and was more accurate witharger display sizes in the moving frame–moving target condi-ion. We speculate that the multiple stationary discs served asreference point, and helped in detecting an incongruent itemithin the moving reference frame. Given the stronger human

bility to perceptually organize coherent movements, as shownn Experiment 2, humans could not ignore the coherent move-

ent of the reference frame, an effect that would be robust evenith larger display sizes.Although Ai’s response time and search slope did not show

vidence of asymmetry reversal, her accuracy was compatibleith the response time performance of Akira and the human

ubjects. Ai’s speed-accuracy trade-off would logically affecter response speed.

. Experiment 3b: asymmetry reversal with frameotion; an additional experiment

In Experiment 3a, one of the two chimpanzees did not show alear tendency to use the coherently moving reference frame inerceptual organization. For further investigation of the naturef chimpanzees’ visual search in a dynamic display, three otherhimpanzees were introduced as subjects. Prior to this experi-ent, they had already learned to detect a moving target among

tationary distractors [28], but were unfamiliar with the oppo-ite task, that of detecting a stationary target among movingistractors, as well as a moving reference frame. In this addi-ional experiment, we tested how they performed in such novelituations.

.1. Methods

.1.1. SubjectsThree chimpanzees, Mari, Pendesa, and Ayumu, participated

n Experiment 3b. Immediately prior to this experiment, they par-icipated in visual search experiments [28], and learned to detectmoving disc target in three conditions in which the target wasefined by movement state (distractors were stationary discs),orm (distractors were moving cross marks), or a conjunction ofhe features (distractors were moving cross marks and stationaryiscs). The first condition was almost the same as the stationaryrame–moving target condition in this study. The three subjectsad no experience detecting a stationary item among movingtems. Pendesa had experienced other visual search tasks thatested discrimination of the perceptual depth of the stimuli [19];he other two new subjects had no other previous visual searchraining.

.1.2. StimuliStimuli and display conditions were the same as in Exper-

ment 3a except for display size, which was fixed to size 6hroughout the sessions.

.1.3. ProcedureA visual search task was presented to the subjects. The proce-

ure was almost the same as in Experiment 3a, although sampleresentation was omitted. After the termination of the warning

lrfle

rain Research 172 (2006) 219–232 229

timulus and the 500 ms interval, a search display (stimuli andhe reference frame) was presented.

The chimpanzees were tested in Experiment 3b with nodvance training or practice. In a 60-trial session, the displayondition was fixed, and the four different display conditionsere tested in four consecutive sessions (one block). The orderf the four sessions was randomly determined. Each chimpanzeeas tested in five blocks (five sessions for each display condi-

ion). Accuracy was analyzed, but because subjects showed noorrect responses in some sessions, correct response time wasot analyzed.

.2. Results

The accuracy data from Experiment 3b are presented in Fig. 7.he marked effect of reference frame movement was apparent

n the stationary target conditions, in which the chimpanzeesade many mistakes with the stationary reference frame, butere very successful with the coherently moving frame. Two-ay ANOVA of reference frame and target conditions found

hat all main effects and the interaction were significant, Ayumu:(1, 4) = 26.01, 512.87, and 33.01 for reference frame condition,

arget condition, and their interaction, respectively, p < 0.01;ari: F(1, 4) = 22.04, 41.36, and 31.38, p < 0.01; Pendesa:

(1, 4) = 10.10, 92.80, and 35.03, p < 0.05. Post-hoc analysesevealed simple main effects of reference frame for the stationaryarget condition, F(1, 8) = 59.03, 50.60, and 39.50 for Ayumu,

eftmost three of the six bars represent the moving target conditions, and theight three of the six bars represent the stationary target conditions. Each bar isor a different subject. Error bars are ±1 S.E. A dashed line indicates the chanceevel (16.6% correct); an open circle indicates the first-session performance ofach subject.

2 ural B

safirmtffjr

6

mppwdInfrgrccsEvtfi

7

hiedpWg(stmpbm

tbadp

[ibhtabfacthttdos

mtoatdskcnal

cicje[cgfctitwovctacmw

30 T. Matsuno, M. Tomonaga / Behavio

These results were not caused by rapid learning to detectpecific targets in the course of the test sessions, as they werepparent even in the first sessions (circles in Fig. 7 indicaterst-session performance). All subjects exhibited the tendencyevealed in the averaged accuracies, although both improve-ent and deterioration were observed in several conditions as

he sessions continued. We suggest that the relatively good per-ormance in the moving frame–stationary target and movingrame–moving target conditions, which were novel for the sub-ects, was the result of generalization from the prior tasks thatequired detecting a moving disc on a neutral gray background.

.3. Discussion

Success in detecting a stationary target depended on theovement of the reference frame. Although subjects had no

rior experience detecting a stationary target and, in fact, theirerformance in the stationary frame–stationary target conditionas below chance levels (16.7%), they were able to successfullyetect targets in the moving frame–stationary target condition.n the same way, success in detecting a moving target was sig-ificantly influenced by the movement of the reference frameor one subject, Mari. These results suggest that subjects moreeadily perceive the “motion” of an absolutely stationary tar-et under the influence of the moving reference frame, althoughelatively good performance in the moving target-moving frameondition suggests that such relative perception did not lead to aomplete perceptual reversal of the movement states. Absolutelytationary discs may serve as a reference point, as discussed inxperiment 3a. In summary, these results further support theiew that chimpanzees partially share with humans the abilityo perceive an object in relation to other objects in the visualeld.

. General discussion

This study investigated visual search in chimpanzees and inumans for moving and stationary targets. The main findingsn this series of experiments were as follows: (1) chimpanzeesxhibited search asymmetry for moving and stationary items asid humans; however, a qualitative difference between chim-anzees and humans was observed in the degree of asymmetry.hile humans found it fairly easy to search for a stationary tar-

et among moving distractors, chimpanzees found it difficult.2) Coherent motion of the distractors facilitated detection of atationary target in humans, but not in chimpanzees. (3) Rela-ive movement states of the stimuli, which were altered by the

ovement of the reference frame, similarly influenced searcherformance in chimpanzees and humans. These results cane discussed from two perspectives: attentional processing ofotion and perceptual organization.First, this study provides further evidence of search asymme-

ry in chimpanzees, and supports shared attentional mechanisms

etween chimpanzees and humans. In Experiment 1, detectingstationary item among moving items was more difficult thanetecting a moving item among stationary items for both chim-anzees and humans. According to previous studies in humans

otcd

rain Research 172 (2006) 219–232

37,50,51], this would suggest that “motion” is a basic featuren the visual system of both species, and the presence of thatasic feature is more easily detected than its absence. Rosen-oltz [38,39] has offered a different account of search asymme-ries using a simple saliency model. According to the model,symmetrical search performance between the conditions cane explained by differences in discriminability of the targetrom distractors in the represented feature space of velocitynd motion direction. In chimpanzees, Tomonaga [46] repli-ated search asymmetry using stationary figures, and confirmedhat the conditions causing search asymmetry are similar forumans and chimpanzees. This study demonstrates that atten-ional processing of the moving items is common, and supportshe view that chimpanzees and humans share the attentional ten-ency to asymmetrically process the visual features dependingn the general presence and absence of features or representedaliency.

From an ecological standpoint, this rapid processing of move-ent against a stationary background is environmentally adap-

ive, and thus, important from the perspective of the evolution ofur visual processing systems. Given its importance, this rapidttentional processing would not be a particular feature of thewo species, but rather would be a general feature in visuallyependent animals. Physiological studies have shown highlyensitive motion detectors in a variety of species, such as mon-eys, birds, and even insects [15], although behavioral evidencesomparing their attentional processing is relatively scarce. Weeed further comparative studies of other species, both distantnd close to humans, to investigate the phylogeny and the evo-ution of attentional processing.

Second, the present study focused on how chimpanzees per-eptually organize moving items. As Gestalt psychologists havendicated [24], coherent motion is a strong determinant of per-eptual grouping for humans. Previous studies with human sub-ects reported that such perceptual organization is helpful forffective visual search performance in a dynamic visual field31,37,55]. The results of the present experiments indicate thathimpanzees have a weaker ability than humans to perceptuallyroup items in their visual field. In Experiment 2, chimpanzeesailed to eliminate the unit of coherently moving items to effi-iently search for a stationary target, and appeared to processhe items one by one, as they did for the non-coherently movingtems. In contrast, humans took advantage of the uniformity ofhe moving items to improve their search rate. This differenceould not have been caused by simple differences in visual res-lution, because one of the subject (Ai) showed almost the sameisual acuity as that of humans in a previous study [30], andhimpanzees detected moving target as efficiently as humans inhis study. Previous studies reported that even macaque monkeysnd pigeons, which are more remotely related to humans thanhimpanzees, can discriminate coherent motion from randomotion as a result of intensive training [3,4,7]. Therefore, whatas difficult for chimpanzees would not have been the detection

f motion coherence but the process of perceptual organizationo put the moving items into one group based on the perceivedoherence and/or to reject them as a unit in search. To efficientlyetect a stationary target among coherently moving distractors,

ural B

wsud

tritddIobimngtrmbtf

uabpseowenpo

isusmwstgcenfsttvtfi

moantdt

idtrsiisicoFphc

A

f1tn(fMhITaaom

R

T. Matsuno, M. Tomonaga / Behavio

e would need two dissociated stages, i.e., the unitization ofpatially separated discs (grouping) and the inhibition of thenit. In the task presented here, the answer to which stage wasifficult for chimpanzees was not evident.

In Experiment 3, chimpanzees showed evidence of an abilityo perceptually organize the coherently moving discs and theeference frame. One plausible explanation for the differencen the results of Experiments 2 and 3 may be the proximity ofhe organized objects. The items to be organized were presentediscretely and scattered within a display in Experiment 2, but theiscs and reference frame were very proximal in Experiment 3.n the latter case, the items appeared to be placed on the surfacef the reference frame; such a strong proximity and associationetween objects may have compensated for the weaker group-ng ability of chimpanzees [12]. Another possible explanation

ay be a difference in the predominant way in which the phe-omena were perceived, which could be as coherently movingrouped items in Experiment 2, and items influenced by rela-ive or induced motion in Experiment 3. The movement of theeference frame could also have altered the perceived relativeovements of the discs. Such relative motion could be perceived

y chimpanzees and influence their performance, even thoughhey could not perceptually group the discs with the referencerame.

These results revealed certain aspects of how chimpanzeesse perception to organize moving visual elements, andddressed some of the differences in perceptual organizationetween chimpanzees and humans. Most previous studies onerceptual organization in non-human animals focused ontationary aspects, such as global/local processing of visuallements and perceptual grouping based on the proximityf elements [5,11,25]. Given the dynamic environment inhich organisms live, however, we should accumulate more

vidence about the dynamic aspects of the perceptual process inon-human animals, such as the perception of relative motion,erceptual grouping of moving objects, and recognition of anbject constructed by interactively moving elements [21,35].

The question regarding the neural basis explaining our behav-oral data that showed similarities and differences between thepecies is difficult to address for two reasons. First, it is stillnclear which neural mechanisms or cortical areas underlieearch asymmetry and perceptual organization. In search asym-etry, bottom–up processes that originate in early visual areasere suggested to explain the phenomenon in a neural model

tudy [26], but no physiological surveys have clarified this pos-ulate. Several studies have claimed that types of perceptualrouping are correlated with neural activity in the primary visualortex [40] or the synchronous neural activity [10], but a gen-ral mechanism explaining perceptual grouping as a whole hasot been elucidated [34]. Second, our knowledge on the brainunctions of chimpanzees is limited. Visual systems in primatepecies exhibit common properties depending on homologies inhe visual cortex [53]. Several studies, however, have reported

he qualitative differences in human brain [6], even in the earlyisual stream [36], in addition to volumetric differences fromhose of other primate species, including chimpanzees. It is dif-cult to discuss the correlation between such differences in brain

rain Research 172 (2006) 219–232 231

icrostructure and those in perception, partly because of the lackf functional studies on the chimpanzees brain that connect brainctivity and behavior. We should look for future development ofon-invasive neuroimaging devices and techniques applicableo chimpanzees. Cumulative efforts to collect behavioral evi-ence would also be helpful for such comparative research onhe evolution of neural systems.

In summary, this study demonstrated not only the similaritiesn the attentional processing of moving objects, but also apparentifferences between chimpanzees and humans in the percep-ual organization of moving items. Presumably, humans haveefined their visual perception abilities from the stage of sen-itivity to visual objects to the stage in which spatially discretetems are relationally perceived in their visual context. Althought would be premature to discuss the evolutionary relevance ofuch a refinement with these few evidences, such an advantagen humans perceptual organization may be related to the largeognitive capacity of humans to recognize variable things atne time, as well as the complex relationships among them.urther experimental studies that extensively compare visualroperties between humans and non-human primates will beelpful in understanding the evolution of visual perception andognition.

cknowledgements

This study was financially supported by Grants-in-Aidor Scientific Research (12002009, 16002001, 13610086 and6300084) and for the 21st Century COE Program (D2) fromhe Ministry of Education, Culture, Sports, Science, and Tech-ology of Japan, and also supported by Research Fellowship16/1060) from the Japan Society for the Promotion of Scienceor Young Scientists. We would like to express my thanks to Dr.

. Tanaka and Dr. T. Matsuzawa of Kyoto University for theirelpful supports and instructions, to Mr. S. Nagumo and Dr. A.zumi of Kyoto University for their technical advices, to Ms.. Imura of Kyoto University for her suggestions on this study,nd to anonymous reviewers for their helpful comments. We arelso grateful to all the staffs at the Primate Research Institutef Kyoto University who work with the chimpanzees for theiranagement of the health of the subjects.

eferences

[1] Asano T, Kojima T, Matsuzawa T, Kubota K, Murofushi K. Objectand color naming in chimpanzees (Pan troglodytes). Proc Jpn Acad1982;58(B):118–22.

[2] Blough DS. Odd-item search in pigeons: display size and transfer effects.J Exp Psychol Anim Behav Process 1989;15:14–22.

[3] Bischof WF, Reid SL, Wylie DRW, Spetch ML. Perception of coher-ent motion in random dot displays by pigeons and humans. PerceptPsychophys 1999;61:1089–101.

[4] Britten KH, Newsome WT, Shadlen MN, Celebrini S, Movshon JA.

A relationship between behavioral choice and the visual responses ofneurons in macaque MT. Vis Neurosci 1996;13:87–100.

[5] Burke D, Everingham P, Rogers T, Hinton M, Hall-Aspland S. Perceptualgrouping in two visually reliant species: human (Homo sapiens) andAustralian sea lions (Neophoca cinerea). Perception 2001;30:1093–106.

2 ural B

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

[

32 T. Matsuno, M. Tomonaga / Behavio

[6] Caceres M, Lachuer J, Zapala MA, Redmond JC, Kudo L, GeschwindDH, et al. Elevated gene expression levels distinguish human from non-human primate brains. Proc Natl Acad Sci USA 2003;100:13030–5.

[7] Cook EP, Maunsell JH. Attentional modulation of behavioral perfor-mance and neuronal responses in middle temporal and ventral intrapari-etal areas of macaque monkey. J Neurosci 2002;22:1994–2004.

[8] Dick M, Ullman S, Sagi D. Parallel and serial processes in motiondetection. Science 1987;237:400–2.

[9] Driver J, McLeod P, Dienes Z. Motion coherence and conjunctionsearch: implications for guided search theory. Percept Psychophys1992;51:79–85.

10] Eckhorn R, Bauer R, Jordan W, Brosch M, Kruse W, Munk M, et al.Coherent oscillations: a mechanism of feature linking in the visual cor-tex? Multiple electrode and correlation analyses in the cat. Biol Cybern1988;60:121–30.

11] Fagot J, Deruelle C. Processing of global and local visual informationand hemispheric specialization in humans (Homo sapiens) and baboons(Papio papio). J Exp Psychol Hum Percept Perform 1997;23:429–42.

12] Fagot J, Tomonaga M. Global and local processing in humans (Homosapiens) and chimpanzees (Pan troglodytes): use of a visual search taskwith compound stimuli. J Comp Psychol 1999;113:3–12.

13] Fagot J, Tomonaga M. Effects of element separation on perceptualgrouping by humans (Homo sapiens) and chimpanzees (Pan troglodytes):perception of Kanizsa illusory figures. Anim Cogn 2001;4:171–7.

14] Grether, Walter F. The magnitude of simultaneous color contrast andsimultaneous brightness contrast for chimpanzee and man. J Exp Psychol1942;30:69–83.

15] Haag J, Denk W, Borst A. Fly motion vision is based on Reichardtdetectors regardless of the signal-to-noise ratio. Proc Natl Acad SciUSA 2004;101:16333–8.

16] Hillstrom AP, Yantis S. Visual motion and attentional capture. PerceptPsychophys 1994;55:399–411.

17] Hopkins WD. Hemispheric specialization for local and global processingof hierarchical visual stimuli in chimpanzees (Pan troglodytes). Neu-ropsychologia 1997;35:343–8.

18] Hopkins WD, Washburn DA. Matching visual stimuli on the basis ofglobal and local features by chimpanzees (Pan troglodytes) and rhesusmonkeys (Macaca mulatta). Anim Cogn 2002;5:27–31.

19] Imura T, Imada M, Tomonaga M. The effect of cast shadows on pictorialdepth perception in chimpanzees and humans. In: Paper presented atthe first international workshop for young psychologists “Evolution anddevelopment of cognition”. 2003.

20] Ivry RB, Cohen A. Asymmetry in visual search for targets defined bydifferences in movement speed. J Exp Psychol Hum Percept Perform1992;18:1045–57.

21] Johansson G. Visual perception of biological motion and a model forits analysis. Percept Psychophys 1973;14:201–11.

23] Kingstone A, Bischof WF. Perceptual grouping and motion coherencein visual search. Psychol Sci 1999;10:151–6.

24] Koffka K. Principles of Gestalt psychology. New York: Harcourt Brace;1935. p. 740.

25] Kuryolo DD, van Nest J, Knepper B. Characteristics of perceptual group-ing in rats. J Comp Psychol 1997;111:126–34.

26] Li Z. Contextual influences in V1 as a basis for pop out and asymmetryin visual search. Proc Natl Acad Sci USA 1999;96:10530–5.

27] Matsuno T, Kawai N, Matsuzawa T. Color classification by chim-panzees (Pan troglodytes) in a matching-to-sample task. Behav BrainRes 2004;148:157–65.

28] Matsuno T, Tomonaga M. Motion-form conjunction search in chim-

panzees (Pan troglodytes). In: Poster presented at the COE21 inter-national symposium “Diversity of cognition, evolution, development,domestication, and pathology”. 2003.

29] Matsuzawa T. Colour naming and classification in a chimpanzee (Pantroglodytes). J Hum Evol 1985;4:283–91.

[

[

rain Research 172 (2006) 219–232

30] Matsuzawa T. Form perception and visual acuity in a chimpanzee. FoliaPrimatol (Basel) 1990;55:24–32.

31] McLeod P, Driver J, Crisp J. Visual search for a conjunction of move-ment and form is parallel. Nature 1988;332:154–5.

32] Nakayama K, Silverman GH. Serial and parallel processing of visualfeature conjunctions. Nature 1986;320:264–5.

33] Ochiai T, Matsuzawa T. Planting trees in an outdoor compound of chim-panzees for an enriched environment. In: Hare V, editor. Proceedingsof the third international conference on environmental enrichment. TheShape of Enrichment. 1997. p. 355–64.

34] Palanca JA, De Angelis GC. Do neuronal synchrony underlie visualfeature grouping? Neuron 2005;46:333–46.

35] Premack D, Premack AJ. Infant attribute value ± to the goal-directedactions of self-propelled objects. J Cogn Neurosci 1997;19:848–56.

36] Preuss TM, Coleman GQ. Human-specific organization of primaryvisual cortex: alternating compartments of dense Cat-301 and calbindinimmunoreactivity in layer 4A. Cereb Cortex 2002;12:671–91.

37] Royden CS, Wolfe JM, Klempen N. Visual search asymmetries inmotion and optic flow fields. Percept Psychophys 2001;63:436–44.

38] Rosenholtz R. A simple saliency model predicts a number of motionpopout phenomena. Vis Res 1999;39:3157–63.

39] Rosenholtz R. Search asymmetries? What search asymmetries? PerceptPsychophys 2001;63:476–89.

40] Sugita Y. Grouping of image fragments in primary visual cortex. Nature1999;401:269–72.

41] Tanaka M. Discrimination and categorization of photographs of naturalobjects by chimpanzees (Pan troglodytes). Anim Cogn 2001;4:201–11.

42] Tomonaga M. A search for search asymmetry in chimpanzees (Pantroglodytes). Percept Mot Skills 1993;76:1287–95.

43] Tomonaga M. Visual search by chimpanzees (Pan), assessment of con-trolling relations. J Exp Anal Behav 1995;63:175–86.

44] Tomonaga M. Transfer of odd-item search performance in a chimpanzee(Pan troglodytes). Percept Mot Skills 1995;80:35–42.

45] Tomonaga M. Visual search for biological motion patterns in chim-panzees (Pan troglodytes). Psychologia 2001;44:46–59.

46] Tomonaga M. Investigating visual perception and cognition in chim-panzees (Pan troglodytes) through visual search and related tasks: frombasic to complex processes. In: Matsuzawa T, editor. Primate origins ofhuman cognition and behavior. Tokyo: Springer; 2001. p. 55–86.

47] Tomonaga M. Action-based distractor effects on the manual responsetimes of chimpanzees during discrimination tasks. Cogn Brain Res2002;13:235–40.

48] Treisman A. Perceptual grouping and attention in visual searchfor features and for objects. J Exp Psychol Hum Percept Perform1982;8:194–214.

49] Treisman A, Gelade G. A feature-integration theory of attention. CognPsychol 1980;12:97–136.

50] Treisman A, Gormican S. Feature analysis in early vision: evidence fromsearch asymmetries. Psychol Rev 1988;95:263–76.

51] Treisman A, Souther J. Search asymmetry: a diagnostic for preattentiveprocessing of separable features. J Exp Psychol Gen 1985;114:285–310.

52] Ushitani T, Fujita K, Sato A. Perceptual organization of motion inpigeons (Columba livia). Jpn Psychol Res 2004;46:170–81.

53] Van Essen D, Van Essen D. Organization of visual areas in macaqueand human cerebral cortex. In: Chalupa LM, Werner JS, editors. Thevisual neurosciences, 1. Cambridge: MIT Press; 2004. p. 445–57.

54] Verghese P, Pelli DG. The information capacity of visual attention. VisRes 1992;32:983–95.

55] Watson DG, Humphreys GW. Segmentation on the basis of linear andlocal rotational motion; motion grouping in visual search. J Exp Psychol

Hum Percept Perform 1999;25:70–82.

56] Williams DW, Sekuler R. Coherent global motion percepts from stochas-tic local motions. Vis Res 1984;24:55–62.

57] Wolfe JM. What can 1 million trials tell us about visual search? PsycholSci 1998;9:33–9.