picture recognition in animals and humans

Behavioural Brain Research 109 (2000) 143165

Review article

Picture recognition in animals and humans

Dalila Bovet a, Jacques Vauclair a,b,*a Centre de recherche en Neurosciences Cogniti6es, CNRS-CRNC, 31, chemin Joseph-Aiguier, 13402 Marseille Cedex 20, France

b Department of Psychology, Uni6ersite de Pro6ence, 29, a6. Robert Schuman, 13621 Aux-en-Pro6ence Cedex 1, France

Received 4 June 1999; received in revised form 27 December 1999; accepted 27 December 1999

Abstract

The question of objectpicture recognition has received relatively little attention in both human and comparative psychology;a paradoxical situation given the important use of image technology (e.g. slides, digitised pictures) made by neuroscientists in theirexperimental investigation of visual cognition. The present review examines the relevant literature pertaining to the question of thecorrespondence between and:or equivalence of real objects and their pictorial representations in animals and humans. Two classesof reactions towards pictures will be considered in turn: acquired responses in picture recognition experiments and spontaneousresponses to pictures of biologically relevant objects (e.g. prey or conspecifics). Our survey will lead to the conclusion that humansshow evidence of picture recognition from an early age; this recognition is, however, facilitated by prior exposure to pictures. Thissame exposure or training effect appears also to be necessary in nonhuman primates as well as in other mammals and in birds.Other factors are also identified as playing a role in the acquired responses to pictures: familiarity with and nature of the stimulusobjects, presence of motion in the image, etc. Spontaneous and adapted reactions to pictures are a wide phenomenon present indifferent phyla including invertebrates but in most instances, this phenomenon is more likely to express confusion between objectsand pictures than discrimination and active correspondence between the two. Finally, given the nature of a picture (e.g.bi-dimensionality, reduction of cues related to depth), it is suggested that objectpicture recognition be envisioned in variouslevels, with true equivalence being a limited case, rarely observed in the behaviour of animals and even humans. 2000 ElsevierScience B.V. All rights reserved.

Keywords: Behaviour; Electrophysiology; Spontaneous responses; Acquired responses; Invertebrates; Birds; Mammals; Humans

www.elsevier.com:locate:bbr

1. Introduction

Researchers in animal and in human cognition fre-quently use photographs or slides in place of realobjects in their studies of categorisation, face recogni-tion, etc., but paradoxically, there are few experimentseither with animals or humans that have explicitlyaddressed the question of the equivalence between anobject and its picture. In other words, it is not obviousthat animal and human subjects do really interpret the2-D stimuli as the 3-D objects they represent. Forexample, the success obtained in training pigeons[40,63] or monkeys [13,80,112] to categorise photo-graphic slides does not prove that the animals under-

stand what the pictures they categorise actuallyrepresent. In fact, as we will see in this paper, somestudies have demonstrated that this is not the case,while others have shown that the establishment of someequivalence between the real object and its pictorialrepresentation is dependent upon both the stimulusdimensions and experimental and:or motivational con-ditions. The present review tries to take stock of thisquestion by examining the available literature for hu-mans (mostly infants) and nonhuman subjects.

This review will first examine experiments concerninghumans and will subsequently consider studies withnonhuman subjects, with the latter being classified intothree categories. The first category comprises of casesof convincing demonstrations in which animals are ableto treat pictures like the stimuli they represent; we canassume that a picture is recognised when animals react

* Corresponding author.E-mail address: [email protected] (J. Vauclair)

0166-4328:00:$ - see front matter 2000 Elsevier Science B.V. All rights reserved.PII: S01 6 6 -4328 (00 )00146 -7

D. Bo6et, J. Vauclair : Beha6ioural Brain Research 109 (2000) 143165144

to a picture as they would react, spontaneously or aftersome training, to the real object. Of course, such reac-tions may vary according to the type of presentedpictures: social behaviour with pictures of conspecifics,fear with threatening stimuli, predator behaviours withpictures of prey, etc.; spontaneous responses and trans-fer of various acquired responses (naming, categorisa-tion, discrimination, cross-modal matching, etc.) inother cases. The second category encompasses the ex-periments which could indicate the existence of picturerecognition but are not really demonstrations becausethe experimental design is questionable (for examplewhen only one subject is involved) or the results (sub-jects preferences or time viewing, discrimination ofindividuals or species, or various spontaneous be-haviours) are not necessarily specifically elicited by thepresented stimuli. The third category includes thoseexperiments which show that animals may havedifficulties with picture recognition.

In addition, two subclasses may be distinguishedamong the studies that have used pictures of living orinanimate objects with animal subjects; the first classreferring to the studies examining learned reactions tostimuli (as it is often the case when the subjects areprimates or birds), while the second class of studiesmeasures spontaneous or natural reactions to the stim-uli (this type of study is frequently seen in experimentsinvolving lower vertebrates or invertebrates). In thislatter case, a very salient feature of the releasing stimu-lus can suffice to induce the reaction. For example, amale redbreast reacts to a lure (e.g. a red tuft of redfeathers) as if it were a real conspecific, even if the luredoes not look like a bird [61]; a colour photograph of amale conspecific may induce a similar reaction but it isnot certain whether, in such cases, the whole stimulushas to be processed and recognised. However, when asubject is trained to respond to real stimuli and thentransfers its response to pictures of those stimuli, or isable to use video images to acquire some informationabout the nature of a real object, this suggests that themost significant features of the pictured stimuli areconsidered and recognised. Therefore, as picture pro-cessing can differ as a function of the kind of response(spontaneous or learned), these two classes will beconsidered separately.

A third classificatory key concerns the issue of thestimuli presented, that is, whether the image is static(photography, slide, digitised picture) or a motion pic-ture, which of course implies some movement and oftensound and may thus greatly facilitate the subjectsreaction to the stimuli. For example, movement is wellknown as releasing predatory behaviour [11], or mayplay an important role in the courtship of many species(see for example Refs. [35,94]).

2. Studies with humans: cross-cultural anddevelopmental studies

Even in humans, recognition of photographs or pic-tures is not as straightforward as it may first appear.Thus, Miller [70] showed that there are interculturaldifferences in picture perception, with humans whohave never seen pictures having difficulty recognisingwhat is represented in black-and-white photographs;this author gives an example originally reported byHerskovits ([41], cited in Miller [70]) who imparts thata Bush Negro woman was initially unable to recognisea photograph of her son until details were pointed outto her. Similarly, Miller cites Kidd [56] who reportedthat Bantus expressed difficulties in recognising objectsin photographs until the details of those objects werehighlighted to them and they then perceived them al-most instantly. Deregowski et al. [29] encountered thesame difficulties with a remote Ethiopian population,but these authors emphasised that members of thispopulation were able to gradually recognise drawingsbut with considerable effort and seemingly finding thetask stressful.

Kennedy [55] has suggested that the subjects studiedby Deregowski and his colleagues could have initiallyrecognised details and then progressively built up acomposite structure of the picture. Miller [70] has alsoshown that even when people recognise objects as rep-resented in pictures, they may experience problemsperceiving the third dimension; thus, depth is often notseen. For example, objects which were represented inthe background appeared to some subjects as beingplaced upon objects represented at the forefront of thepicture. Miller concluded that the insight that pro-duces an overcoming of flatness cues may require verylittle experience with pictures, but experience in perceiv-ing objects in the three-dimensional world is not suffi-cient to perceive those objects in pictorialrepresentations and that direct experience with picturesmight also be necessary for the perception of depth cuesin pictorial materials ([70], p. 148). Deregowski [28]also relied on cross-cultural studies to understand themechanisms of the perception and representation ofspace; coming to the conclusion that even if peopleunfamiliar with 2-D representations had moredifficulties with picture recognition, the same kind ofdifficulties could occur in pictorial and nonpictorialcultures.

An interesting experiment has presented results thatcontradict the preceding conclusions. Hochberg andBrooks [44] studied the behaviour of a child brought upuntil the age of 19 months without being exposed topictures; when tested at the end of this period, this childwas able to recognise and to name various objects fromhis familiar environment represented in photographsand line drawings. However, this childs environment

D. Bo6et, J. Vauclair : Beha6ioural Brain Research 109 (2000) 143165 145

was not totally devoid of pictorial representations (it isimpossible in the USA!), as he accidentally saw repre-sentations of objects, for example, billboards on thehighway.

Even very early in life, infants are already able todiscriminate between 2-D and 3-D stimuli. Thus, Pippand Haith [76] observed that 4-week-old infants differ-entiated 3-D from 2-D forms, with the 2-D formsinvoking shorter fixation times. This experiment wasrepeated with 8-week-old infants who also processed a3-D form in a different manner than a 2-D form, not interms of overall fixation time but in the visual scanpatterns elicited; there were more eye movements inresponse to 3-D forms. An experiment carried out byAppel and Campos [1] may offer some insight into howinfants differentiate 2-D and 3-D stimuli; resultsshowed that 8-week-old infants could discriminate be-tween stimuli differing only in terms of binocular dis-parity, that is, when they were habituated tostereograms without retinal disparity and then pre-sented with the same stereogram with retinal disparity,heart rate increased indicating that they dishabituated.

Bower [6] studied the reaction of neonates when areal ball or its colour photograph was presented at sucha distance that the ball could only be touched but nograsped. The real object was frequently contacted butthe photograph not at all (it did not even elicit handraising) although the infants attentively stared at it. Incontrast, very young infants (under 23 days of age)were shown to perform a similar amount of reaching inthe presence of either a three-dimensional graspableobject (an orange textured sphere) or a two-dimen-sional picture of it [32]. Slater et al. [91] showed thatnew-borns (mean age: 2 days and 21 h) could discrimi-nate real objects from their photographs; all partici-pants looked longer at the real objects, even withmonocular viewing, leading the authors to suggest thatmotion parallax was a salient cue for this discrimina-tion. Interestingly, there was no evidence that the new-borns were able to recognise stimulus similarity acrossdimensions, thus, for these new-borns, differences be-tween objects and their two-dimensional representa-tions seemed to be more detectable or salient than theirsimilarities.

Other investigations indicate that human babies arenot only able, with very little or no experience withphotographs, to discriminate these from real objects,but also to recognise what they represent. In an experi-ment carried out by Rose [81], 6-month-old infantspresented with various geometric stimuli were not onlyable, in an habituation and visual preference test, todiscriminate 2-D from 3-D stimuli, but also to transferhabituation from 2-D to similar 3-D stimuli, or in-versely from 3-D to similar 2-D objects. This apparentease of processing objects and pictures in a similar wayis not, however, a consistently reported result. For

example, some authors found that very young children(less than 30 months of age) did not interpret thepictures as representing current reality; when the loca-tion of a hidden toy was demonstrated using photo-graphic stimuli, 24-month-olds did not use thisinformation to retrieve the toy, a task which 30-month-olds readily performed [23].

Dirks and Gibson [30] have shown that 5-month-oldinfants, without any experience of photographs, whohad previously been habituated to an unfamiliar, liveface, showed no change in fixation time when presentedwith a slide of the same person, but dishabituated whenpresented with a slide of a novel person who differed insex, hair colour, and hairstyle from the familiar face.However, if the novel person was of the same sex, andhad the same hair colour and hairstyle as the familiarface, no difference in fixation time was observed, sug-gesting that these infants could see the similarity be-tween a live person and their photograph using rathergross physiognomic features. Similar findings are re-ported in a study by Barrera and Maurer [2] who foundthat 3-month-old babies who had never seen photo-graphs looked longer at their mothers photograph thanat a strangers one. Related evidence of an early sensi-tivity of infants to pictures of conspecifics comes fromthe study of the phenomenon of gaze following; forexample, Hood et al. [46] showed that 3-month-oldscould detect another individuals gaze shifts even whenpresented as digitised pictures of adult faces.

Cross-modal experiments can also be valuable in thestudy of transfer from objects to pictures or the reverse.Rose et al. [82] showed that such a transfer was possiblewith 12-month-old infants, but that it depended onfamiliarisation time: with a familiarisation time of 30 s,infants were only able to perform a cross-modal trans-fer from touch to vision (real objects), and to transferin a visualvisual task from real objects to both theiroutline drawings and coloured silhouettes. With anincreased familiarisation time of 45 s, subjects were ableto cross-modally transfer from touch to real objects, totheir outline drawings and to their coloured silhouettes.However, with a familiarisation time of 15 s, infantswere no longer able to transfer in a visualvisual taskfrom real objects to their outline drawings, or colouredsilhouettes.

Streri and Molina [92] conducted another experimenton cross-modal transfer with infants of only 2 monthsof age. Somewhat paradoxically, these authors foundthat pictures were more easily recognised than realobjects in a transfer from vision to touch: transferoccurred between felt objects and their 2-D visual sil-houettes, but not between felt objects and their visualcounterparts. The authors hypothesised that this rela-tive ease could be explained by the fact that infants takemore information from seeing stimuli than from touch-ing them. Thus, the use of pictures could have sim-


plified and reduced visual information, consequentlyrendering the object more easily identifiable for theinfant who had only tactile experience of the object.Subsequent experiments performed by the same authorsusing the habituation procedure provided some evi-dence that 2-month-old babies were able to perceiveboth the commonalties and the differences of 3-D ob-jects and their 2-D representations.

Experiments measuring event-related potentials havereached the same conclusion as most of the experimentspresented above. Recordings of ERPs in 7-month-oldbabies while they were watching pictures of facesshowed that the active components differentiated be-tween a happy face and a fearful face but not betweenan angry face and a fearful face [71], suggesting that theinfants recognised what was presented on the pictures.

Perhaps the most impressive demonstration of younginfants abilities in interpreting pictorial representationscan be found in a study on intermodal transfer betweenoral exploration of objects and visual matching. Ineffect, in two experiments, Kaye and Bower [52]showed that new-borns as young as 12 h old were ableto match tactile shapes (pacifiers) with visual represen-tations of the pacifiers shafts displayed as digitisedcoloured or black-and-white images on a computerscreen.

In summary, cross-cultural studies have demon-strated that adults who have never seen any two-dimen-sional representations may experience difficultiesrecognising pictures; these participants need some ex-planation and some experience with a photograph (or adrawing) before being able to perceive what it repre-sents. However, developmental studies reveal that theability to recognise significant information in picturessuch as photographs is evident even in very earlyinfancy (demonstrated at 3 months or even younger).This apparent paradox between adult and infant perfor-mance will be discussed in the conclusion of this article.Human babies are also able to discriminate real objectsfrom their pictorial representations; this precociousability does not, however, preclude infants and eventoddlers from confusing an object and its referents. Themethods, populations and main results of the studiesdiscussed above are summarised in Table 1.

3. Studies with animals: convincing demonstrations

3.1. Spontaneous responses to pictures

In this section, studies that provide rather unambigu-ous evidence of spontaneous responses to pictures as ifthey were real objects are considered; studies usingstatic and motion pictures will be presented in turn.

3.1.1. Responses to still picturesThe perception of still pictures can elicit adapted

responses by monkeys. Thus, von Heusser [43] reportedthat a tamed marmoset displayed grabbing responses infront of photographs representing different prey (e.g.ants, butterflies) but displayed fear reactions whenshown a picture of a cobra. In a similar vein, Rosenfeldand van Hoesen [83] related that naive rhesus monkeysreacted with hasty retreat, threat responses and vocali-sations to the first presentation of slides of rhesusmonkeys faces. These monkeys also displayed abortiveapproachretreat before touching the stimuli (i.e. theslide projector), this behaviour did not persist, however,because the subjects quickly realised that the stimuliwere only pictures (i.e. unresponsive, immobile stimuli).Comparable findings were obtained with cynomolgusmacaques by Kyes et al. [60], namely, that dominantmonkeys produced threatening gestures when shown,for example, pictures of gorillas or humans, while sub-ordinates gave submissive responses to these samestimuli.

Sackett [87,88] presented real-life-size coloured slidesof various social stimuli to rhesus monkeys; subjectsexhibited different responses to the various stimuli,many of these responses being appropriate to the situa-tion. Moreover, the level of responses to a given slidevaried according to age and rearing conditions: subjectsreared in isolation showed more exploration of non-monkey pictures and pictures without any social com-munication content than with socially relevant pictures.Results of another experiment [59] with hamadryasbaboons suggest that these monkeys were able to recog-nise slides of conspecifics. Subjects were given controlover slide selection and viewing time, with some slidesdepicting individual troop members and others depict-ing various facial areas of a troop member; the baboonswere highly reliable in their choices, with dominancestatus seemingly a primary factor in troop memberpreference and slides of full faces being consistentlychosen and the eye region attracting the greatestattention.

Overman and Doty [72] have investigated hemi-spheric specialisation for face processing in pigtailmacaques. Prior to testing, the authors examined iftheir subjects would react in a similar manner to realmodels and pictures of humans and monkeys; variousemotional reactions were measured (e.g. vocalisations,lip-smacking, etc.) at the first presentation of differentcategories of slides and results showed that pictures ofhumans and monkeys were clearly differentiated fromother classes of stimuli such as flowers, insects orlandscapes.

The available neurophysiological evidence supportsthe view that nonhuman primates establish some corre-spondence between photographs and the individualmonkeys they represent. For instance, studies with


macaques have revealed that groups of neurones intheir left inferotemporal cortex are responsive to facesof other monkeys and are sensitive to the identity of themonkey [74,113]. It occurred that similar groups of

neurones were firing in response to life facial stimuli, topicture stimuli and to still video [75].

In addition to nonhuman primates, it seems thatsheep may also be good candidates for studying picture

Table 1Studies with humans

Task Nature of pictures Age Results Reference

Increased heart rate in theStereogram or simple Eight weeks Appel andDiscrimination of stimulipicture of a color drawingwith binocular disparity Campos [1]dishabituation phase

from stimuli withoutbinocular disparity

Mothers photograph Barrera andPreference for mothersColor slides Three monthspicture Maurer [2]recognition

Bower [6]NeonatesColour photographs Hand raising is elicited by aBehavioural observationsreal ball but not by itsphotographERP shows differenceSix months Nelson and deTwo-thirds-size digitizedEvent-related potentials

Haan [71]colour photographsrecorded from babies between observation of themothers face and awatching pictures of theirstrangers face, but lookingmothers face or a

strangers face time does notFailure to find the hiddenTwenty-four and 30 monthsPhotographsFinding a hidden object when Deloache and

the location is Burns [23]object in 24-month-oldchildren but not indemonstrated with

photographs 30-month-oldTrying to grasp the depicted Deloache et al.Color photographsBehavioural observations Nine months

[25]objects, despitediscriminating betweenobjects and picturesSame types of difficulty Deregowski [28]Adults and children with orBlack-and-whiteCross-cultural studies ofoccur in pictorial andhuman picture perception photographs and drawings without experiencenonpictorial cultures

Line drawings Adults without experiencePresentation of pictures Difficulty recognizing what Deregowski eta picture represents al. [29]

Life-size colour slides Perception of similarityHabituation to a live face, Dirks andFive monthsthen measurement of between a live person and Gibson [30]

their photographfixation times for a slide ofthe same and a novel face

Color photographs Similar amount of reachingTwenty-three days Dodwell et al.Comparison of amount ofwith a real ball and with itsreaching to a real ball and [32]

to its picture picturePresentation of pictures Adults without experience Herskovits [41]Difficulty recognizing whatBlack-and-white

a picture representsphotographsHochberg andNaming the represented Photographs and line Nineteen months without Correct naming

experience Brooks [44]drawingsobjectsThree months Correct detection Hood et al. [46]Colour digitized picturesDetection of an adults

change of gaze directionIntermodal transfer occursTwelve hours Kaye andDigitized colored orIntermodal transfer between

oral and visual exploration black-and-white images Bower [52]of objects

Kidd [56]Presentation of pictures Black-and-white Difficulty recognizing whatAdults without experiencephotographs a picture represents

Adults and children withBlack-and-whiteReview of cross-cultural Difficulty in perception and Miller [70]recognition of various typesresearch more or less experiencephotographs and drawingsof pictures

Recording of event-related Nelson and deEvent-related potentials varyBlack-and-white slides Seven monthsHaan [71]potentials with the expression of the

faces presentedSimple black-and-white Four weeks and 8 weeksDiscrimination between 2-D Pipp and HaithVisual behavior

[76]differentiates between 2-Dsilhouettesand 3-D stimuliand 3-D stimuli


Table 1 (Continued)

Age ResultsNature of pictures ReferenceTask

Six months Rose [81]Babies are able to discrim-Discrimination between 3-D Black-and-whitephotographs inate between 3-D andand 2-D stimuli, and transfer

2-D stimuli, but also toof habituation from 2-D tosimilar 3-D stimuli, or perceive similarity across

dimensionsinversely from 3-D to similar2-D objects

Cross-modal transfer isCross-modal transfer between Colored outline drawings Rose et al. [82]Twelve monthsand silhouettestouched objects and their possible but depends on

familiarization timepicturesTwo days and 21 hours Neonates discriminateDiscrimination between 3-D Slater et al. [91]Black-and-white

between 3-D and 2-D stimuli,photographsand 2-D stimuli, and transferbut do not perceive similarityof habituation from 2-D to

similar 3-D stimuli, or across dimensionsinversely from 3-D to similar2-D objects

SilhouettesCross-modal transfer from Two months Pictures are more easily Streri and Molinavision to touch recognized than real objects [92]

in this kind of cross-modaltransfer

recognition in animals. Thus, a study by Vandenheedeand Bouissou [101] indicates that sheep recognised a 2-Dstimulus at its first presentation. In this study, the fearreactions of ewes were tested when the subjects wereseparately presented with a full-size slide of a human, asheep, or a control stimulus (a traffic cone); the ewesshowed reduced fear reactions in the presence of a sheepsphotograph, as with real conspecifics and, moreover,sniffing was primarily directed towards the anogenitalregion and the head, which corresponds to behavioursdirected towards real conspecifics. However, the humanslide failed to induce fear reactions, as they occurred witha real human or even a human-like model [101], thussuggesting the possibility that recognition of 2-D stimulicould be easier when stimuli are conspecifics. A subse-quent experiment [4] showed that, in ewes, a slide of anunknown individual of its own breed significantly re-duced fear reactions compared to a slide of an unknownindividual of a different breed, this latter result suggest-ing that the ewes can recognise the characteristics of theirbreed on the slide.

Finally, Clun Forest and Dalesbred sheep showed anability to discriminate black-and-white photographs de-picting faces of sheep versus human faces when they weretested with a procedure of spontaneous choice in aY-maze [54]. In addition, the same study demonstratedthat Clun ewes could also distinguish between the facesof male and female breed members and between a troughwith and without food. Moreover, some subjects wereobserved displaying overt social reactions toward stimuli,such as licking the pictures (C. Fabre-Nys, pers. com-mun., May 1998).

3.1.2. Reactions to motion picturesPlimpton et al. [77] showed social stimuli (for example

a threatening male stimulus) to juvenile bonnet macaques

via colour videotape recordings and observed thesesubjects in the presence of their mother; they exhibitedappropriate responses depending on the nature of thesocial display, that is, they behaved submissively towardthe threatening male and searched for contact with theirmother while they approached a passive female. Herzogand Hopf [42] showed different colour films to wild-bornand laboratory-born squirrel monkeys. While the presen-tation of predators (cats, snakes or avian predators)caused specific alarm and flight reactions, the subjects didnot emit any alarm response when nonpredator mam-mals were shown. Further, they reacted in the same wayas they did in real situations upon seeing preparation offood or insects walking and also reacted to films ofhuman beings as to real people. These subjects demon-strated face recognition; upon seeing in the film acaretaker who had recently removed a dead neonate, thesquirrel monkeys behaved as if they were facing a realterrestrial predator. No difference was shown betweenwild and laboratory born-subjects.

In studies with birds, the use of predators picturesalso yielded positive indications of picture recognition.For example, in a study by Evans and Marler [34]which used video images as stimuli, domestic cockerelswere shown to respond with similar alarm calls inresponse to an aerial predator model when either videosof hens or real caged hens were present. The use ofsocial stimuli, notably in tasks requiring the recognitionof conspecifics, can provide important insight on theability of birds to match objects with their pictures.Shimizu [90] observed that when video images of fe-males were presented to male pigeons, the duration ofthe males social display was not significantly differentfrom that which they performed in front of the livebird. When video images of another bird (a cockatoo)


or of an empty room were presented to the subjects,they showed much shorter, or no, display. Finally, theduration of the display was longer when video imageswere in motion rather than still and when only the headregion was visible rather then when only the bodyregion was visible.

An experiment carried out by McQuoid and Galef[67] with juvenile Burmese fowls provides evidence thatobserving feeding among conspecifics via video tapeshas similar effects as observing live conspecifics onsubsequent feeding behaviour (i.e. preferences for fooddishes); thus, the sight of a video of a conspecificfeeding before testing, even without sound, reduced thelatency of the first peck. Further, a videotape of a fowlactually feeding on a food dish was more effective inreducing the latency of the first peck and in enhancingpreference for that type of food dish than a videotapesimply representing a fowl (either active or immobile)near a food dish.

Jenssen [49] studied the recognition of motion pic-tures by lizards. In this study, female lizards couldchoose between two films, one with males displayingthe normal courtship behaviour and the other withmales presenting altered displays; in most cases, thefemales chose the normal displays. This experimentsuggests that female lizards could recognise male lizardsin the film and that they could be sensitive to theirbehaviour. When video sequences displaying aggressivedisplays of their own species were shown to malelizards, these animals responded with the appropriateresponses (e.g. head-bobbing, crest erection) which theywould exhibit in front of live opponents. Moreover,such behaviours were inhibited when video-recordedsequences of heterospecific lizards were presented [65].

Some experiments have reported evidence of imagerecognition in fishes, at least when video images (imply-ing motion) were used. Thus, Rowland [85] showed thatmale and female sticklebacks reacted to video images ofa zigzag dancing male played at a normal or slightlyfaster tempo in a similar way as they would react to alive male; when the tempo was slower or much fasterthan the normal tempo, the animals were less attracted.In guppies, consistency of mate preference was studiedby presenting females with males under three experi-mental procedures: live males behind clear glass, livemales behind one-way glass and images of males onvideotapes. Females spent significantly more time inproximity to males behind clear glass than in video andone-way glass presentations, but they spent equal timewith males behind one-way glass and videotaped males[58]. In all three experiments, females responded tostimuli by displaying sexually oriented behaviours andthe results show that when interaction was not possible,a videotaped male was as attractive as a real male.

Some studies suggest that even invertebrates canrecognise video images. For example, Clark and Uetz

[11] carried out an experiment with jumping spiders andfound that in a V-maze choice, spiders preferentiallychose a videotape with moving prey to a videotapewithout prey. In addition, the spiders did not discrimi-nate between a live prey and its simultaneously pre-sented video image and they behaved in a mannercomparable to their reactions with life stimuli whenthey were presented with televised images of prey in-sects (attack), conspecifics (courtship) and heterospe-cific spider species (retreat).

3.2. Acquired responses in picture recognition

We will now turn to a consideration of studies thathave attempted to test the abilities of different animalspecies using pictures (still and in motion) as stimuli.Although most of the experiments summarised belowwere aimed at investigating picture recognition, wehave also considered experiments for which the explicitgoal was not picture recognition as these studies alsopresent findings relevant to this topic.

3.2.1. Reactions to still picturesHayes and Hayes [39] reared a female chimpanzee,

Vicki, in their home, almost like a human child. Whilethey did not specifically train her in picture perception,they did test her ability to recognise pictures depicted inbooks and other materials and to imitate actions illus-trated in films, photographs, and line drawings. Vickiwas able to recognise most of the pictures she saw, evenwhen the pictures were presented as black-and-whitedrawings. Nevertheless, those authors reported thatVicki did not confuse photographs with real objects;Vicki did not try to grasp 2-D objects and when shepointed for example to pictures of beverages, she saidcup, and let the person who was with her go to thekitchen and get a drink. Gardner and Gardner [38]showed that four chimpanzees that had been famil-iarised with pictures could also recognise and name inAmerican Sign Language various objects representedon new slides.

Sarah, an adult chimpanzee experienced with filmsand photographs was shown videotaped scenes of ahuman actor struggling with one of eight problems andthen presented with two photographs that could consti-tute a solution to the problem [78]. Sarah chose thecorrect photograph on seven of the eight problemssuggesting that she recognised both what happened inthe films and the objects as depicted in photographsand as films. Furthermore, Savage-Rumbaugh et al.[89] trained chimpanzees to categorise various objectsinto the two categories tools or foods; when theobjects were replaced by photographs (the subjects werealready familiarised with pictures), the two chimpanzeestested were still able to categorise (and thus to recog-nise) them.


An interesting study on the recognition of pictures ofindividual faces by chimpanzees was carried out byBauer and Philip [3]. The authors found that threechimpanzees initially trained to match photographs offaces of the same individuals were able to later matchdifferent vocalisations to facial portraits of the familiarindividuals.

Experiments on cross-modal perception can alsoprovide us with useful indications concerning the ques-tion of picture recognition in nonhuman primates. Forexample, Davenport and Rogers [16] trained three apes,one orang-utan and two chimpanzees, who were unfa-miliar with photographs, to match a visual sample (thereal object) to an haptically presented object. When thevisual sample was a photograph, subjects responseswere clearly above chance (80% or more) from thebeginning. Further, there was little difference in accu-racy between colour and black-and-white photographs,and no difference between real objects and their colourphotographs. Thus, it seems that apes are able, evenwithout any familiarity with photographs, to recognisethem and treat them like real objects. In another exper-iment [17], five chimpanzees trained only in a haptic tovisual cross-modal experiment with real objects weresubmitted to the same sort of problems with photo-graphs or drawings; four of the subjects performedsignificantly above chance with full-size colour photo-graphs, with full-size or half-size black-and-white pho-tographs and with full-size line drawings. However,cross modal-perception was better with photographsthan with line drawings.

An experiment of cross-modal matching of objectswith their photographs was conducted by Malone et al.[66] with two adolescent male rhesus monkeys. One ofthe monkeys was trained on visual to haptic matchingto sample, and the other on haptic to visual matchingto sample (the visual stimuli were full-sized colourphotographs of the haptic objects). While both mon-keys succeeded on the task, they needed prior trainingto do so and the authors highlighted that it was unclearwhether training was necessary because subjects haddifficulty mastering the matching-to-sample procedure,or if they had to learn first the equivalence betweenphotographs and objects. In a subsequent experiment[97], the same subjects were required to perform thesame task but with black-and-white photographs, sil-houette photographs, and outlines drawings of the ob-jects, that is, forms of stimuli with which they had hadno prior training; the monkeys were still able to per-form the task when visual stimuli were black-and-whitephotographs and silhouette photographs, but not withoutline drawings.

Zimmermann and Hochberg [114] trained infant rhe-sus monkeys (5150 days of age) to discriminate be-tween flat and solid objects (e.g. squares and cubes) andthen tested the transfer to photographs and outlines

figures of these objects. The results showed that thesesubjects were able to make consistent responses topictorial representations of the stimuli (photographs ordrawings) and that while the presence of shadow facili-tated transfer, this feature was not necessary.

Dasser [14] showed that two Java monkeys were ableafter a few trials, to identify novel views (full face orfull animal) of a familiar conspecific presented onslides, and that another subject could match differentbody parts of the same familiar group members. In asubsequent experiment, Dasser [15] showed that twoadult female Java monkeys, first familiarised with slidesof their conspecifics, were able to identify motheroff-spring pairs or to match views of offspring to theirmother, a task which requires the recognition of groupmembers on the slides.

In a recent study by our group [5], olive baboonswere trained on the natural category of food versusnon-food with real objects. After categorical transferwith novel items, subjects were trained again with onepair of cut-out pictures each of which belonged to thetwo previously learned categories; after this limitedtraining, categorical transfer was high in both baboonsfor cut-out photos of the food and non-food objects.Results of the experiment and of additional controlsituations involving various modes of picture presenta-tions further demonstrated the abilities of the baboonsto relate real objects to their pictorial representations.

A consideration of studies using nonprimates revealsthat most of these experiments have been conductedwith birds, especially pigeons, but studies with otherzoological groups will be considered subsequently. Oneof the first studies concerned with the question ofpicture recognition in animals addressed the ability ofpigeons to recognise a conspecific shown in a picture[62]. The authors used three pigeons with experimentalhistories of attacking a mirror target, but not pictures,and submitted them to an intermittent schedule ofreinforcement for key pecking; they found comparableresults for both temporal pattern and locus of attacks(the head region) to that reported in studies with live,taxidermally stuffed pigeons or mirror targets. Subse-quent experiments demonstrated that an upright silhou-ette, white-on-black silhouette of a pigeon, with orwithout eye, was more effective in controlling attackthan an inverted silhouette, an outline of a pigeon, or apiece of coloured paper.

A study of transfer of discrimination from solidobjects to pictures by pigeons was carried out by Cabe[10]. The pigeons, which were naive with respect topictorial stimuli, were first trained to discriminate twoobjects, and then four groups of four pigeons weretested in reinforcement reversal with objects, black-and-white photographs, silhouettes or line drawings of thoseobjects; negative transfer was expected in all cases if thepigeons recognised the pictures. In fact, negative trans-


fer occurred with objects (for all four pigeons), withphotographs (for three of the four pigeons) and withsilhouettes (for all four pigeons), but not with linedrawings (none of the four pigeons). Four birds testedonly with objects were then re-tested without reversalusing black-and-white photographs and showed a posi-tive transfer and, in addition, subsequent training todiscriminate objects from their photographs showedthat this was fairly easily obtained. These results sug-gest that the birds did not confound objects and theirpictures and, altogether, these experiments also attestthat prior experience is not necessary for picturerecognition.

Delius [26] trained eight pigeons to discriminatespherical objects from nonspherical objects and follow-ing this training, seven birds out of eight were able totransfer to black-and-white photographs, colour photo-graphs, or drawings of these objects, as well as tophotographs of novel objects; the overall transfer toblack-and-white photographs was best, that to draw-ings intermediate, and that to colour photographsworst. The author suggested that the pigeons did worstwith colour photographs because their colour vision isat least pentachromatic and the colour photographywas matched to the trichromatic colour vision of hu-mans. Watanabe [105] trained 12 experimentally naivepigeons to discriminate, for one half, between realobjects (edible and nonedible objects) and their photo-graphs, and for the remaining half between food andnon-food objects in the same set of stimuli (real objectsand their photographs). The pigeons performed high oneither task and showed generalisation to novel stimulifor the two tasks, hence demonstrating that pigeonswere able to treat photographs like real objects, and todiscriminate between them, according to their trainingand categorical judgement.

Jitsumori and Ohkubo [51] found that four experi-mentally naive pigeons trained to discriminate right-side-up and upside-down orientations of colour slidesof natural scenes depicting humans, transferred thisdiscrimination to new slides of the same kind. Both theorientations of the human figures and of the back-ground scenes controlled this discrimination, but whenthese slides were oriented in the opposite direction, thebackground orientation cue was the dominant feature.The birds were also able to categorise by orientationnatural objects (humans, apes, monkeys and birds) on awhite background, indicating that the subjects recog-nised the objects presented.

3.2.2. Reactions to motion picturesInfant chimpanzees [68] and even baboons [102] were

able, after limited experience, to match what they ob-served on a television screen to events occurring else-where in order to determine the location of a hiddengoal object in a familiar outdoor field. Along similar

lines of research, Menzel et al. [69] showed that twochimpanzees could use mirrors or live video images tomove their hands in the appropriate direction and makecontact with target (food) objects.

It has been demonstrated that fear reactions can alsobe learned by use of videotaped demonstrators. Forexample, naive rhesus monkeys acquired a fear ofsnakes through watching videotapes of conspecifics re-acting fearfully to snakes; note however, that the mon-keys did not acquire a fear of flowers through watchingvideotapes of monkeys reacting fearfully to flowers [12].

To summarise, it appears that the experiments pre-sented in this section reveal that picture recognition ispossible by animals, even without previous experience.Thus, spontaneous adapted responses were displayed tosignificant stimuli photographs (prey, predators, orconspecifics) by monkeys. Other mammals (sheep)showed adapted responses to slides of conspecifics andsimilar responses to pictures of conspecifics were alsoevident in other species (birds, lizards, fishes and evenin some invertebrates) when the pictures were presentedin motion; such spontaneous responses to these stimulidid not require any previous training with pictures.

Apes, monkeys and pigeons were also able to transferacquired responses from objects to pictures in varioustasks, however, it is difficult to know in such caseswhether familiarisation with the pictures is necessaryfor their recognition. The methods, species and mainfindings of the papers that have reported some evidenceof picture recognition in animals are outlined in Table2.

4. Studies with animals: experiments that could indicatepicture recognition

This section reviews those experiments which providesome cues indicating the presence of picture recognitionbut that may not constitute real proof of such anability.

4.1. Spontaneous responses to pictures

As in the previous section, studies that used still andmotion pictures will be considered in turn.

4.1.1. Reactions to still picturesHumphrey [47] used visual stimuli, either plain fields

of light colour, or photographs or films, and a simplechoice procedure with two adolescent rhesus monkeys;subjects could push two buttons to choose one of thetwo stimuli presented on a screen. Humphrey inter-preted rhesus monkeys preferences in terms of interest(determined by the information content in the stimuli)and pleasure (determined by features such as colourand brightness), with these two factors determining the


Table 2Convincing demonstrations of picture recognition in animals

Task ResultsNature of pictures ReferenceSpecies

Black-and-white photographs Chimpanzee Correct matching Bauer and PhilipMatching vocalizations to facialportraits of familiar conspe- [3]cifics

Behavioural observations Sheep Appropriate spontaneous re-Life-size colour slides Bouissou et al. [4]sponses to images of conspe-cifics

Colour photographsCategorization of objects and Olive baboons Correct transfer occurs from Bovet and Vau-objects to pictures clair [5]their pictures into food and

non-food categoriesDiscrimination transfer from Black-and-white photographs or Pigeon Correct transfer occurs from Cabe [10]

objects to picturesdrawingstwo objects to their picturesBehavioural observations Appropriate spontaneous re-Black-and-white video images Jumping spider Clark and Uetz

[11]sponses to images of conspe-cifics, prey and heterospecificspiders

Colour video images Rhesus monkey Fear of snakes acquired but notAcquisition of fear of snakes or Cook and Minekaof flowers by observation of fear of flowers [12]conspecifics

Long-tailedMatching various novel views of Colour slides Correct matching Dasser [14]macaquea conspecific

Colour slides Long-tailed Dasser [15]Categorization of motheroff- Correct categorization ormatchingspring pairs or matching macaque

mother to offspringMatching a touched but unseen Correct cross-modal matching Davenport andChimpanzeeBlack-and-white and colour pho-

tographsobject to its photograph Rogers [16]Black-and-white and colour pho-Matching a touched but unseen Chimpanzee Davenport et al.Correct cross-modal matching

object to various 2-D repre- tographs (life-size and half-size) [17]and line drawingssentationsBlack-and-white and colour pho- Delius [26]Categorization of objects and Pigeon Correct transfer occurs from

objects to picturestographs or drawingspictures as spherical and non-spherical objects

Colour video imagesBehavioural observations Appropriate spontaneous re-Domestic Evans and Marler[34]sponses to predatorscockerel

ChimpanzeeNaming the represented objects Correct namingColour slides Gardner andin American Sign Language Gardner [38]

Black-and-white films, photo- ChimpanzeeMatching pictures and imitating Correct matching and imitation Hayes and Hayes[39]actions illustrated in pictures graphs, and line drawings

Colour films Squirrel monkey Appropriate spontaneous re-Behavioural observations Herzog and Hopf[42]sponses to predators, food, and

humansLizard Female choice of normal maleKodachrome II indoor films (BW Jenssen [49]Behavioural observations

displaysor NB unspecified)Colour slidesDiscrimination of right-side-up Pigeon Correct discrimination transfer Jitsumori and

and upside-down orientations Ohkubo [51]to novel viewsof scenes

Life-size black-and-white photo-Spontaneous choice in a Y- Sheep Discrimination between human Kendrick et al.maze between pictures and sheep faces, between malegraphs [54]

and females conspecifics andbetween a trough with andwithout foodAppropriate spontaneous re-Guppy Kodric-BrownLife-size colour video imagesBehavioural observationssponses to images of conspe- and Nicoletto [58]cifics

HamadryasColour slidesSpontaneous choice between Choices consistent with social Kyes and Cand-context land [59]slides of conspecifics baboon

Long-tailedColour slidesBehavioural observations Appropriate spontaneous re- Kyes et al. [60]macaque sponses to pictures of gorillas

and humansColour photographs and draw-Reinforced attack of a pigeon Looney andAttack of the target picturePigeon

target Cohen [62]comparable to attack of a liveingstarget


Table 2 (Continued)

Task SpeciesNature of pictures Results Reference

LizardBehavioural observations Appropriate spontaneous re-Life-size colour video images Macedonia et al.[65]sponses to images of conspecifics

Cross-modal matching of objects Full-size colour photographs Rhesus monkey Correct cross-modal matching Malone et al. [66]and their photographs

Burmese fowlBehavioural observations Acquisition of preferences for McQuoid andLife-size colour video imagesfood dishes by observation of Galef [67]conspecifics

Finding a hidden object with Menzel et al. [68]Black-and-white video images Chimpanzee Correct transfer from video im-ages to real situationvideo images used to demon-

strate locationMenzel et al. [69]Moving their hands to make Colour live video images Chimpanzee Correct hand movements

contact with target objectsshown on video

Pigtail macaqueBehavioural observations Spontaneous choice and appro- Overman and DotyColour slidespriate spontaneous responses to [72]slides of humans and conspecifics

Patterson-Kane etColour video imagesDiscrimination transfer between Domestic hen Transfer occurs only when thetwo stimuli to discriminate havevarious hens and objects to al. [73]different colourstheir pictures

Drawings and still video images Rhesus monkey Similar neuronal responses toElectrophysiological recording Perrett et al. [75]images as to life facial stimuli

Bonnet macaque Appropriate spontaneous re-Colour video imagesBehavioural observations Plimpton et al. [77]sponses to pictures of conspecifics

Premack andPhotographs and film ChimpanzeeChoice of a photograph as the Correct choicesolution to a problem presented Woodruff [78]in a film

Discriminations of conspecifics Slides Rosenfeld and vanRhesus monkey Correct discrimination and ap-propriate spontaneous responsesfaces and behavioural observa- Hoesen [83]to conspecifics facestions

Stickleback Appropriate spontaneous re-Colour video images Rowland [85]Behavioural observationssponses to images of conspecifics

Colour slidesMeasurement of visual and Rhesus monkey Sackett [87]Visual and tactile responses varyconsistently with the nature oftactile responsesstimuli, and with the subjectsage and experience

Rhesus monkeyBehavioural observations Appropriate spontaneous re-Colour slides Sackett [88]sponses to pictures of conspecifics

Savage-RumbaughCategorization of objects and Photographs Chimpanzee Correct transfer occurs fromet al. [89]their pictures into food and objects to pictures

non-food categoriesPigeon Courtship display to images ofMotion and still colour video im-Behavioural observations Shimizu [90]

ages conspecificsMatching an object touched but Black-and-white and colour full- Correct cross-modal matching Tolan et al. [97]Rhesus monkey

size photographs, silhouettes andunseen to various 2-D for photographs and silhouettes,but not for outline drawingsoutline drawingsrepresentations

Behavioural observations Colour slides Sheep Reduction of fear and appropri- Vandenheede andBouissou [100]ate spontaneous responses to im-

ages of conspecificsVauclair [102]Colour video images Guinea baboon Correct transfer from video im-Finding a hidden object when

ages to real situationthe location is demonstratedusing video

MarmosetBehavioural observations Appropriate spontaneous re-Photographs von Heusser [43]sponses to prey and predatorsCorrect transfer from objects to Watanabe [105]Colour slidesCategorization of objects and Pigeon

their pictures into food and pictures, and correct discrimina-non-food categories, and dis- tion between objects and picturescrimination between objectsand pictures

Black-and-white photographs or Zimmermann andRhesus monkeyDiscrimination transfer between Discrimination transfer occurssquares and cubes to their drawings Hochberg [114]pictures


strength and direction of preferences; when the twofactors are set against each other, interest overrodepleasure in determining the preference. When naivemonkeys were tested for preferences for coloured pho-tographs paired with plain fields of light, they firstshowed negative preferences and signs of fear. How-ever, as they became more experienced, the signs of feardropped away and they showed positive preferences forcolour photographs (the same pattern of responsechange was observed when films were introduced in-stead of photographs). The photographs were dividedinto six classes, and the rank order of preferencesexhibited by the monkeys was other animals, mon-keys, men, flowers, abstract paintings and foods.In a further experiment, Humphrey [48] used noveltypreference for slides to investigate how rhesus monkeyscould differentiate between individual animals of thesame species; the finding was that monkeys to whichdomestic animals were unfamiliar treated individualdomestic animals of the same species as being closelysimilar, but treated individual monkeys as being differ-ent from each other. However, monkeys who had beenexposed for 6 months to many pictures of animals,treated all individuals as different from each other.Demaria and Thierry [27] conducted a rather similarexperiment with female stumptailed macaques. Thesefemales were submitted to slides displaying individualprimates or non-primates and the results showed thatsubjects looked longer at slides of individuals of theirown species than at slides depicting other macaquesspecies; moreover, they looked more at adult femalescarrying infants than at adult females alone. Withpictures of non-primates animals, subjects looked mostat slides of felids. However, spontaneous social re-sponses, like facial expressions, were very rare.

Fox [36] observed the responses of young and adultdogs (the breed was not given) presented for the firsttime with a life-size dog painting. The young dogssniffed more at hind leg and inguinal regions, whileadults sniffed more at ear and anal areas; as all theseareas are normally investigated by conspecifics, we caninfer that the subjects had perceived the correspondencebetween the painting and a real dog.

In sheep, Kendrick and Baldwin [53] recorded re-sponses of cells in the temporal cortex of awake sub-jects and demonstrated that some of these specificallyresponded to slides (photographs and drawings) offaces (but not to upside-down faces or profiles). More-over, different groups of these cells were influenced byrelevant social factors, such as dominance, breed, famil-iarity, and facial expression. Franklin and Hutson [37]investigated the reactions of sheep to full-size colourphotographs of one sheep and to colour films of mov-ing sheep and found that the sheep reacted to 2-Dimages as if they were real animals: subjects were slowto approach a sheep facing them, but they approached

without hesitation or followed a sheep displayed inprofile; reactions which were heightened when the im-age was moving. The most attractive stimulus was thefilm of sheep moving across the screen towards the exit.

Several studies are available concerning the ability ofbirds to spontaneously display adapted responses topictures of biologically relevant situations. For exam-ple, dark-eyed juncos which could choose betweenslides of their winter and summer habitats spent moretime in front of the pictures that were consistent withtheir season of capture and laboratory photoperiodconditions compared to the inconsistent habitat [79].Klopfer [57] performed imprinting tests with Pekingducklings, using various decoys or images and demon-strated that the ducklings followed both decoys andimages. However, the images elicited responses thatwere not strictly equivalent to the three-dimensionaldecoys; in effect, the ducklings reacted differently todifferent decoys (according to their colours, and ac-cording to those they were accustomed to follow) butnot to different two-dimensional representations ofducks. One aspect of the task is particularly interestingin this experiment: the comparison between decoys andimages, because the only difference between those tworepresentations is the presence or absence of the thirddimension, that is, the lack of three-dimensionality wassufficient to cause a decrease in attention or in re-sponses to feature differences.

4.1.2. Reactions to motion picturesRosenthal et al. [84] presented green swordtail fe-

males with video-recorded sequences of the same malewhich was either engaged in an active courtship display,or which performed similar levels of feeding activity, orwhich remained inactive (control sequences showedfood particles in movement or an empty aquarium).Female behaviour patterns differentiated between thepre-stimulus, stimulus and post-stimulus periods for thethree stimuli showing a male, but not for the twocontrols; courtship displays elicited more activity thanany other stimulus, and there were no significant differ-ences between the responses to the feeding and inactivesequences.

4.2. Acquired responses in picture recognition

We will now examine studies of acquired reactionswhich fall into the category of responses that couldpossibly be evidence for the ability of picture recogni-tion in animals.

In an experiment conducted by Tomonaga et al. [98],a sample of students and a language trained femalechimpanzee (called Ai) were trained to recognise videostill pictures of individual faces of humans or chim-panzees, presented at various orientations. The experi-ment yielded two main findings indicative of picture


recognition in the chimpanzee; firstly, it was moredifficult for Ai to recognise human faces than chim-panzees faces, while the opposite was shown for hu-man subjects (they had more difficulty recognisingchimpanzees faces than human faces) and, secondly, itwas more difficult for her to recognise inverted faces orhorizontal faces than to recognise upright faces (asimilar, but more pronounced effect being obtainedwith human participants). It must be noted that experi-ments conducted with macaque monkeys [8,83] failed toshow this inversion effect, but this could be due todifferences in processing rotated complex visual stimuli.For example, monkeys and some apes could be moresuited to doing this type of rotation because they live inan environment in which they often hang upside-down[103], the above findings do not, therefore, necessarilyimply that subjects were unable to recognise the pic-tures presented. Nevertheless, Swartz [93] showed, witha visual fixation habituationdishabituation paradigm,that infant pigtail macaques (3 months old) could dis-criminate between colour photographs of faces of threemacaque species (pigtail, cynomolgus, and stumptailed)when they were presented upright, but not when thefaces were displayed upside-down.

Other experiments using schematic drawings of mon-keys bodies indicate that longtailed macaques wereable to discriminate one monkey from other monkeys,basing their recognition on the limited informationprovided by the black-and-white shape and texture oftheir body characteristics [31]. Such findings, however,do not present any obvious interpretation regardingpicture recognition. Firstly, as stated by the author ofthe above study it remains questionable whether themonkey has a knowledge of the representational natureof the image (Dittrich [31], p. 150). Secondly, we knowfrom other studies that monkeys can correctly cate-gorise images of different classes of stimuli by usingsome absolute cues which are not constitutive of the tobe categorised stimuli. For example, DAmato and vanSant [13] have shown that their monkeys used anirrelevant red patch to form the person versus non-per-son category. Thus, caution is in order before conclud-ing that seemingly appropriate classification skills meanthat the animal realises the relationship between apicture and the real object.

Watanabe and Ito [106] trained pigeons to discrimi-nate between colour slides of two pigeons faces. Whilediscrimination was apparently effortless for two stimulieasily discriminated by human observers, it was difficultwith the stimuli which humans also struggled to differ-entiate. When the S stimulus was replaced by itsscrambled parts, subjects did not respond; such a reac-tion seems to indicate that the birds recognised that thestimuli depicted on the slides represented conspecifics(which could be recognised only when the faces werenot scrambled).

Lumsden [64] conducted an experiment with onepigeon; the bird was trained to discriminate one geo-metric object from two others, after which transfer wasexamined when the object, its cut-out photograph, orits line drawing was shown at various orientations.Response curves were the same for photographs andfor three-dimensional objects: generalisation was goodat 0, 45, and 135, poor at 180, and was absent for 90.Although the line drawings were responded to at thelowest rate, the pattern of responding was similar. In asubsequent experiment, the pigeon was trained to dis-criminate between the object and its photograph dis-played at 45: the bird then generalised thatdiscrimination to photos presented at other orienta-tions. We should note however that there was only onesubject involved in this experiment.

Wilkie et al. [108] trained four pigeons to discrimi-nate between pictures taken in the vicinity of the loft towhich they had been raised and other areas they hadnot visited, with four other pigeons that were nottrained to home being tested as a control. The resultsshowed that after training with only eight slides, bothgroups were able to transfer and then to discriminatethe two categories of slides, but homing pigeons werebetter than nonhoming pigeons. In the same paper, theauthors mention the experiment of Honig and Ouellette[45], in which eight pigeons were taught to discriminatecolour pictures of various views of two ends of a longroom. Following this task, the pigeons had to discrimi-nate between the two ends of the real room; a feederwas placed at each end of the of the test room but onlyone feeder contained food: for the congruent group, itwas in the same location that had been positive duringthe previous slide discrimination procedure, while forthe incongruent group, it was the opposite end of theroom. The subjects in the congruent condition tookconsistently less time to find the correct feeder than theincongruent subjects. In a similar study, Wilkie et al.cited an unpublished study by Willson et al. [109] inwhich eight pigeons were placed outside the laboratoryfor 20 min prior to each training session in a picturediscrimination task; for four of the pigeons the placethey had seen outside was presented as the positivestimuli (relevant place), whereas for the other fourpigeons, the visited place was not pictured at all (irrel-evant place). The relevant place birds acquired thediscrimination more quickly than did the irrelevantplace subjects. Such experiments suggest that pigeonscan perceive the correspondence between pictorial stim-uli and the place they represent.

The above findings provide further information re-garding the issue of picture recognition. In some exper-iments, animals showed differential preferences forpictures and appeared to be able to discriminate be-tween them. However, they did not treat them as theconspecifics they represented: for example macaques


did not display social behaviour toward pictures ofconspecifics [27,47] and ducklings did not react to thepictures in exactly the same way as they did with decoys[57]. In other experiments, behaviours were indicativeof picture discrimination [79,84] but were not specifi-cally directed toward the stimuli. While visual experi-ence with real locations can facilitate the discriminationof photographs by pigeons, and vice versa, it is not yetclear how this facilitation occurs.

Table 3 summarises the methods, species and mainresults of the studies that could be indicative of picturerecognition in animal species.

5. Difficulty with picture recognition

This section reviews those studies which show adifficulty or failure of the subject to react to 2-D stimulias if they were meaningful or 3-D stimuli.

5.1. Spontaneous responses

It is important to note that socially salient stimulipresented as pictures do not always elicit overt re-sponses in birds or even in monkeys. Butler andWoolpy [9] studied visual attention in rhesus monkeyssubmitted to various slides or motion pictures of otherrhesus monkeys but their results are not easy to inter-pret because they appear to be quite contradictory; theamount of visual attention given to slides of conspe-cifics was not different from attention devoted to anhomogeneous illuminated screen. Such a result seems toimply that the monkeys did not recognise the slide asrepresenting one of their conspecifics although viewingbehaviour (and thus attention) was more important (i.e.monkeys looked longer) when motion pictures wereprojected in the normal orientation than when theywere projected upside-down.

Table 3Experiments which could indicate picture recognition in animals

Task Nature of pictures Species Results Reference

Dog Fox [36]PaintingsBehavioural observations Sniffing the areas normally investi-gated on conspecifics

Life-size colour slides and Franklin andBehavioural observations Sheep Following images of conspecificsfilms Hutson [37]

Transfer occursColour slides Honig andPigeonTransfer of discrimination of twoOuellette [45]ends of a room from pictures to

real placesConsistent choice Humphrey [47]Choice between two visual stimuli Rhesus monkeyColour slides and films

(plain field or various stimuli)Humphrey [48]Discrimination of individuals de-Rhesus monkeyNovelty preference Slides

pends on subjects experienceSpecifics cells in the temporal cor-Electrophysiological recording Kendrick andSheepColour slides and black-and-

white drawings tex respond to faces Baldwin [53]Colour filmBehavioural observations Pekin duckling Spontaneous following of moving Klopfer [57]

ducksTransfer of discrimination between PigeonCut-out photograph and line Same type of curve depending on Lumsden [64]

drawing object orientation, but overall levelobjects at various orientationsof responding considerably lessto pictures

Measurement of time spent in Roberts andDark-eyed junco Spontaneous choice of appropriateColour slideshabitats according to the seasonfront of slides of habitats Weigl [79]

Measurement of female interest Colour video images Green swordtail Frequency of behavior patterns in Rosenthal et al.females depends on the male [84]courtship

Swartz [93]Colour photographsDiscrimination between faces of Pigtail macaque Difficulty recognizing invertedthree macaque species faces

Discrimination between faces of Tomonaga etColour video still pictures Difficulty recognizing invertedChimpanzeeal. [98]faceschimpanzees and humans indi-

vidualsWilkie et al.Colour slidesDiscrimination between pictures of Pigeon Discrimination between pictures

facilitated by prior experience withlocations [108]the locationDiscrimination between picturesPigeonColour slidesDiscrimination between pictures of Willson et al.

[109]facilitated by prior experience withlocationsthe location


Social recognition experiments which used pictorialstimuli with hens, failed to indicate any transfer ofdiscrimination from live birds to photographs [19]: henspreferred flock-mates rather than unfamiliar conspe-cifics (even when they saw only their heads and necks)when presented with live stimuli, but they failed toshow any preference with photographic stimuli. Similarfindings were observed in a study of hens shown videosequences [21]: hens neither took longer to eat nearunfamiliar conspecifics than near flockmates, nor nearhigh-ranking flockmates than near low-ranking flock-mates, as they usually did when they saw live stimulibehind a clear screen. Pigeons also failed to exhibit anynatural social response when they were presented withlife-size moving video images of conspecifics [86].

5.2. Acquired responses

5.2.1. Reactions to still picturesWinner and Ettlinger [110] trained two chimpanzees

with no prior experience with photographs in a match-ing-to-sample task. First, the subjects had to match realobjects to real objects and subsequently had to matchobjects with their photographs. Initially, they were un-able to perform the task successfully with performanceremaining at chance levels for the first 4 days and thenrising moderately, but not consistently, above chance.The second experiment of Winner and Ettlinger at-tempted to replicate the results obtained by Davenportet al. [17] (see above); in their experiment, two chim-panzees with no prior experience with photographswere required to transfer a discrimination between pairsof objects that were felt but not seen to their photo-graphic representations or vice versa. The new subjectsresponded significantly above chance when tested onlywith objects and at chance level when required totransfer a learned response from a felt object to aphotograph or from a photograph to a felt object. Theauthors suggest that in Davenport et al.s experimentthe objects were not paired by size: consequently, sub-jects might have succeeded by choosing the bigger one.

Jitsumori [50] has also demonstrated that picturerecognition is difficult for untrained animals. The taskconsisted of training four monkeys (two of them hadprior experience with discrimination problems betweenpictures containing or not containing monkeys) andfour experimentally naive pigeons to discriminate be-tween normally oriented displays and topbottom re-versals. If the monkeys saw meaningful objects in theseslides, then transfer was supposed to occur with variousnovel slides; subjects were trained with a go:no-godiscrimination task with colour pictures of full humans,and then tested with other pictures of humans, mon-keys, birds, mammals and man-made objects. Bothmonkeys and pigeons showed good transfer to novel

human pictures but when tested with other pictures,levels of performance revealed considerable interindi-vidual variation, namely, in pigeons and in nonexperi-enced monkeys transfer was relatively good for somepictures but not for others. Thus, the overall perfor-mance was inconsistent and successful transfer might beexplained by perceptual similarities among the slidespresented in a fixed orientation. In this study, only oneof the experienced monkeys produced results suggestiveof the perception of meaningful objects in pictures.

Another experiment [19] attempted to establishwhether or not transfer between geographical locationsand photographs of those locations occurred in homingpigeons. Eight pigeons were trained to discriminatephotographs of two geographical locations, havingbeen given visual experience of a real geographicallocation beforehand. Half of the birds were transportedto one of the two locations that appeared in the photo-graphs, while the remaining subjects were transportedto a third, irrelevant location. Although there was nosignificant difference in acquisition or transfer to novelstimuli between the two groups, the authors suggestthat this might be due to their methods (inadequateamount of experience outside or lack of immediatereward for learning about the environment), but also todifferences between human and bird vision (see above).Moreover, it is possible that the pigeons were process-ing the far-distance views and the near-distance views indifferent manners.

A study with laying hens by Bradshaw and Dawkins[7] attempted to replicate the experiment performed byDasser [14] (see above) with macaques. Hens weretrained to discriminate between slides of either familiaror unfamiliar conspecifics and were then presented withnovel views of these birds; during training, the right-hand side of a hens head was presented, whereas thenovel stimulus set was composed of pictures of left-hand side of the corresponding hens head, a frontalview, or a view of the tail or feet. The birds failed togeneralise discrimination from training slides (both fa-miliar and unfamiliar) to novel view categories and theauthors concluded that their study provided no evi-dence that the hens perceived the slides presented asrepresentations of their group members. Ryan and Lea[86] obtained somewhat comparable results in a studyin which pigeons and chickens were trained to discrimi-nate between slides of two individuals (two pigeons ortwo chickens). For both species, the chicken slides werelearned faster and better than the pigeon slides, withthe pigeons performances being much worse thanchickens on both chicken and pigeon stimuli. More-over, only one pigeon out of six was able to discrimi-nate slides of pigeons, and none learned to discriminatebetween two different stuffed pigeons, even though asubsequent experiment proved that they readily dis-criminated individual live pigeons.


Table 4Experiments showing difficulties recognizing pictures in animals

Nature of picturesTask Species Results Reference

Color slides Bradshaw andDiscrimination not facilitated byLaying henDiscrimination between picturesfamiliarityof familiar or unfamiliar con- Dawkins [7]

specificsBlack-and-white and colour Butler andBehavioural observations Rhesus monkey No spontaneous responses; noslides and motion pictures more attention to slides of con- Woolpy [9]

specifics than to a homoge-neously illuminated screen

Spontaneous discrimination be- Dawkins [18]Discrimination occurs for liveDomestic hensLife-size colour photographstween familiar and unfamiliar hens but not for photographs of

hensconspecifics, either live or pre-sented on photographs

Dawkins et al.Discrimination between pictures Colour slides Discrimination between picturesPigeonof locations [19]of locations is not facilitated by

experienceDEath andLife-size colour video sequences Domestic hensSpontaneous discrimination be- Discrimination of live hens but

tween familiar and unfamiliar Dawkins [21]not for those presented on videoconspecifics, either live or pre-sented on video

Jitsumori [50]Colour slides Difficulty transferring discrimi-Monkey and pi-Discrimination of right-side-upand upside-down orientations geon nation to various classes ofof scenes slides

Transfer occurs only when theDiscrimination transfer between Patterson-Kane etColour video images Domestic hentwo stimuli to discriminate havevarious hens and objects to al. [73]different colourstheir pictures

Colour slides and moving videoDiscrimination of individual pi- Ryan and LeaPigeon and Great difficulty in identifying[86]chickengeons and chickens and be- novel views of an individual, noimages

spontaneous responseshavioural observationsCategorization of objects and Watanabe [104]Correct transfer from objects toPigeonColour slides

their pictures into food and pictures occurs only for a natu-non-food categories or into ral category (food)arbitrary categories

Matching an object touched but Chimpanzee Failure to match objects withBlack-and-white and colour full- Winner andEttlinger [110]unseen to its photograph their photographssize photographs

A study by Watanabe [105] is particularly interestingin the discussion of picture recognition because it sug-gests that objectpicture equivalence can be performedrelatively easily when there is some functional basis.Twenty-four pigeons were divided into four experimen-tal groups: two object-to-picture groups and two pic-ture-to-object groups; one of the object-to-picturegroups and one of the picture-to-object groups weretrained on a natural concept (food objects were S forhalf, and non-food objects were S , and it was theopposite for the remaining half) while the other twogroups were trained on a pseudoconcept (an arbitrarygrouping of edible and nonedible objects as positive andnegative stimuli). When tested with the natural concept,the subjects showed a good transfer of discrimination inboth object-to-picture and picture-to-object conditions,but no transfer was observed with the pseudoconcept.Such a result indicates that picture recognition candepend on the consistency of the task.

5.2.2. Reactions to motion pictureAttempts to train domestic hens to transfer from real

stimuli to video images generally produced negativeresults, although, depending on experimental condi-tions, the birds could use some features of the patterns(e.g. the colour) in their discrimination [73]. It wasconcluded from this study that complex video images,such as those required to recognise social stimuli, arenot equivalent to the real stimuli. In addition, somepigeons did not transfer a learned discrimination fromlive conspecifics to their photographs and had greatdifficulties in discriminating between slides of individu-als (although they easily discriminated live conspecifics).

The results of the studies reported in this section aresomewhat contradictory to the findings summarised inSections 3 and 4. In effect, they demonstrate that picturerecognition in animals is not obvious and is dependenton experimental factors. In several experiments, mon-keys and birds (such as pigeons and chicken) failed todisplay an interest in photographs of conspecifics.Moreover, different tasks involving picture recognitionhave reported a failure to demonstrate such an ability;thus chimpanzees failed to realise a cross modal match-ing and only one monkey (out of four), which was


already familiarised with photographs, was able to dis-criminate the orientation of photographs. In addition,it was shown that experience with a particular place didnot facilitate the discrimination between that locationand another. Finally, it was shown that when testedwith a natural concept (food), pigeons transferred adiscrimination from object to picture and vice versa,but they did not demonstrate such transfer when theywere tested with an arbitrary pseudoconcept. In brief,the experiments summarised in this section highlight theimportance of biological relevance in picture recogni-tion tasks.

Table 4 presents an outline of the investigations thatdemonstrated a failure to recognise pictures in animals.

6. Conclusion

One of the main conclusions of this survey is thatvisual stimuli presented as pictures (either as black-and-white photos, colour photos, slides, or videos) are notnecessarily immediately recognised by non-human andeven human subjects. In this final section, we willsummarise the principal results concerning picturerecognition in the much-studied species and attempt toidentify some of the factors which may be responsiblefor the apparent difficulties in recognising pictorialstimuli. Finally, we will suggest some possible stepswhich could be useful in describing picture processing;ranging from feature discrimination, to correspondence,and ultimately, to strict equivalence between a 3-Dobject and its pictorial representation and a consider-ation of the issue of confusion between pictures and theobjects they represent.

6.1. Summary of main findings

Very young humans appear to be able to recognisephotographs from 2 or 3 months of age and at an evenearlier age they are likely to discriminate real objectsfrom their pictorial representations. Paradoxically, pic-ture recognition seems to present greater difficulties foradults who are unaccustomed to seeing photographsand drawings. Thus, it could be hypothesised that suchan ability is innate but that this ability diminishes if theperson has grown up without opportunities to see 2-Drepresentations. In this case (which is less and less likelyto happen in contemporary societies), such individualsare accustomed to seeing the world of objects aroundwhich are inevitably characterised by the presence offeatures like colours, depth, and motion parallax; whenpresented with pictures, these people, who have lived inan exclusively 3-D environment, would experiencedifficulties and some familiarity and:or training withpictures would then be required in order to recognisethe stimuli which lack those features.

The available literature is quite convincing concern-ing the abilities of several animal species familiar withpictures to recognise such 2-D stimuli. In mammals ingeneral (but with most evidence coming from experi-ments conducted with monkeys and apes), it also seemsthat picture recognition is possible for both adults andyoung even if the animals have never been exposed toany picture prior to the experiment. In this latter case,recognition seems to be more difficult and appears todepend on the nature of stimuli and on the experimen-tal conditions (see below). Some experiments, in whichsubjects have to transfer what they have learned withreal objects to the pictures of these objects (e.g. incross-modal transfer between touch and vision), haveshown that these tasks present serious difficulties forthe subjects. Furthermore, because some training isoften necessary in order to perform picture recognitiontasks, the training phase probably allows subjects to getfamiliarised with pictures prior to testing; this require-ment implies that subjects are rarely naive with respectto viewing pictures or their discrimination.

The studies reporting spontaneous responses to pic-tures are interesting because they can provide usefulindications on the perceptual and cognitive processesinvolved in picture recognition performed by trulynaive subjects. These studies have shown that monkeysand other mammals (sheep and dogs) can, at first sight,adaptively respond to various animals or foods (al-though pictures of conspecifics seem to be respondedmore easily to than pictures of other categories ofstimuli) presented on slides. We can speculate that thisease presumably expresses the fact that these animalsconfuse the real objects and their pictures; nevertheless,this recognition can be quite precise, if we consider thatsome animals are able (e.g. [4]) to differentiate individu-als from their own breed from individuals belonging toother breeds. It is also worth noting that transfer isgenerally better for pictorial stimuli which better match(at least for a human viewpoint) real objects, that is,motion films are more easily recognised than still pic-tures, slides are better recognised that colour photo-graphs, the latter leading to better performancecompared to black-and-white photographs and linedrawings.

It may be an interesting observation that in the firstsection of our review (convincing demonstrations),studies concerning mammals are more numerous thanstudies concerning birds, and that the opposite is trueof the third section (difficulties with picture recogni-tion). Actually, the pattern of results obtained withbirds is quite different compared to other zoologicalgroups; a divergence which may be explained in part bythe fact that bird vision is different from mammalvision. With regard to pictures, photographs used inexperiments with these animals are usually matched forhuman vision and lack some critical features of birds


vision, such as UV light [111], and offer false colourrepresentations for these subjects because the stimuliare based on the trichromatic colour vision of humansand not on the pentachromatic vision of birds (e.g.[33]). Motion pictures are also matched for humanvision. In motion pictures, still images are shown insuccession to produce the impression of a continuousmoving image. But the frequency at which a flickeringstimulus starts to appear continuous is higher in birds(notably in pigeons and in chicken) than in humans[20,107]. Further, chickens are myopic (they can recog-nise a conspecific only at a very short distance [18]) andit is possible that they process far-distance views in amanner distinct from that used for near-distance views.The above considerations may thus explain some of theproblems encountered by birds in interpreting picturesand why, for example, transfer is sometimes better forblack-and-white photographs than for colour photo-graphs. Nevertheless, some of the experiments reportedearlier showed a good transfer from objects to picturesor confusion between the slides or films and the animalsthey represented.

Finally, a handful of studies have convincinglydemonstrated that responses to pictures are not limitedto birds and mammals; reptiles, fishes and even inverte-brates reacted strongly to video images depicting bio-logically significant stimuli (conspecifics, prey orpredators, for examples). It is, however, not surprisingthat animals of different phyla respond to salient visualcues in similar ways, i.e. as they would respond to realobjects; for many years, experimental ethologys tech-niques have used visual lures for identifying the stimu-lus characteristics of social and aggressive behaviour inanimals (e.g. the pioneering and now classic studiesemploying cardboard models of a Herring gulls headby Tinbergen and Perdeck, [95]).

6.2. Factors influencing picture recognition

The experiments reported in the section on acquiredresponses to pictures fail to provide a clear and defini-tive answer to the question of capacities of differentanimal species for processing pictorial representation ofobjects. Given that the most advanced and detailedstudies have been conducted with birds, it may beuseful to list some of the factors that authors havehighlighted as playing a determining role in the extentand limits of picture recognition. Some of these factorshave been summarised in the discussion of dEath andDawkins [21] article reporting a failure of domestichens to discriminate between familiar and unfamiliarconspecifics on videos and are thoroughly described inthe review by dEath [20]. A first and obvious factor hasto be mentioned: pictures, being still or in motion, area