The Philosophical Case Against Visual Images

Home , Imagery, Mental image

Center for Cognitive Science University of NewSouth Wales P.O. Box 1, Kensington 2033 NSW.,Australia [email protected] s.unsw.oz.au New Focus for Imagery Debate "crucial experiment", since a falsified prediction can always be blamedon one of the auxiliary hypotheses on In their study of reasoning with diagrammatic and non- which any theory depends. Nonetheless, the text-book dia~,rammatic representations, Larkin and Simon(1987) exampleof a "crucial experiment" is the null result of are concerned with external representations and the Michelson-Morleytest of the speed of light, and I explicitly avoid drawinginferences about the bearing of take the null results of our owninvestigations to show their work on the issue of internal, mental that, like the luminiferous ether, the pictorial medium representations. Nonetheless, we mayinfer the bearing does not exist. It is in this sense that the possibility of of their work on internal representations from the reinterpreting visual patterns in mental imagery has theories of Kosslyn, Finke and other ’pictorialists’ who recently emerged as new focus for the long-standing take internal representations to be importantly like controversy. The question of whether, and under what external ones regarding their ’privileged’ spatial conditions, novel information maybe discovered from properties of depicting and resembling their referents. images may shed new light on the debate, since it Thus, Finke (1990) suggests that "perceptual provides a new meansfor testing the properties of the interpretive processes are applied to mental images in conjectured pictorial medium.andthe claimed parallel much the same way that they are applied to actual between imagery and perception. physical objects. In this sense, imaginedobjects can be "interpreted" muchlike physical objects" (1990, p. 18). Elsewhere he suggests that "The image discoveries "Equivalence" of Imagery & Perception which then ’emerge’ resemble the way perceptual Specifically, the possibility of reinterpreting an image discoveries can follow the active exploration and follows as a direct implication of the pictorial theory manipulation of physical objects" (1990, p. 171). which posits an "equivalence" between imagery and After twenty years of controversy, the ’imagery perception. On the pictorial view, a mental image is debate’ concerning the ’format’ of visual mental conceivedto be a "surrogate percept, allowing people to representations is widely regarded as having become detect somepattern or property in a rememberedscene stalled and the impasse has even led some (Anderson that they did not encode explicitly whenthey saw the 1978) to conclude that the issue between pictorialism scene initially" (Pinker and Finke 1980, p. 246). It and the ’tacit knowledge’alternative is undecidable in in this sense that the uninterpreted images in a spatial principle on the basis of behavioral evidence. mediumare themselves supposed to be "functionally At least part of the reason for the persistence of the equivalent to physical objects or events" (Finke, 1980 imagery debate has been the fact that the dispute has p. 113), and cause the same mechanismsto be activated centered upon alternative explanations of the samebody as in actual visual perception itself (ibid, p. 130). of chronometric evidence. The debate has come to Kosslyn (1987, p. 149) explains, one purpose appear intractable because the two contending theories imagery involves "recognition processes" to discover make identical predictions for chronometric evidence information which is not stored explicitly in memory and, accordingly, adducing new evidence of time- and thus we "look" at our images in a way which is dependentmeasures, as has repeatedly been done, cannot analogous to the way we look at external objects in strengthen the case for a pictorial, spatial medium order to inspect them. against the rival tacit knowledge theory. Thus, experiments are needed on which the contending Divergent Predictions accountsdeliver different predictions. Our ownevidence concerns perceptual organization By assimilating imagery so closely with vision, indeed tasks which provide unequivocal criteria of the by claiming their "equivalence", pictorialism is successful rotation, inspection and re-interpretation of committed to predicting closely similar "perceptual" images using "recognition processes" and "shape phenomenain imagery to those found in perception classification" procedures. Despite the demonstratedease itself. It is this deep commitmentto the perceptual of our task under perceptual conditions, naive subjects character of imagery which is the source of its have generally been unable to succeed in the tasks under vulnerability to such asymmetriesas those of Chambers imagery conditions as would be predicted on the and Reisberg (1985) and our ownresults. pictorial theory. Thus, Kosslynsuggests that "imageinterpretation is Of course, as historians and philosophers of science at the heart of the role of imagery in cognition (if one well know, there can be strictly no such thing as a cannot inspect imaged patterns, they are useless)"

12 (1988, p. 249). He says "The recognition mechanisms Finke and Slayton (1988) have extended this work, [of vision] can be usedin imageryas a wayof accessing providingfurther evidence"that people are capableof stored information"(1988, p. 264)and his modelentails makingunexpected discoveries in imagery" and that that "imagesdepict visual information, and that this novel patterns can "emerge" from within imaged informationis interpreted by someof the samesorts of patterns. classificatory proceduresused in classifying sensory input during vision" (1980, p. 32). Kosslynexplains Response by Finke, Pinker and Farah. further that "the purposesof imagery,in large part, parallel those of vision" and "onemay ’recognize’ parts Most recently, Finke, Pinker and Farah (1989) have and properties of imagedobjects that had not been soughtto reinforce these claims with newexperiments previouslyconsidexed" through the "use of recognition which also purport to showthat subjects can inspect processes" (1987, p. 149). The significance and reinterpret their images by "applying shape reinterpreting imagesas a crucial test of the pictorial classification proceduresto the informationin imagery" theoryis evidentin Kosslyn’sexplicit predictions: (1989,p. 51). This latter workis of particular interest becauseit has beenspecifically designedto counter the Theimage is formedby forcing a changeof state in skeptical conclusionswarranted by the negativeresults the visual buffer in the attended region, whichcan of Chambersand Reisberg and sets the scene for our then be reprocessedas if it wereperceptual input (e.g., owninvestigation. Thus,it is in the light of this clash the shape could be recategorized), thereby of experimentalresults and theoretical claims that our accomplishingthe purposesof imagerythat parallel own experiments are to be understood: Our new those of perception.(1987, p. 155; emphasisadded) experimentsavoid the specific objections by Finke et al. and by falsifying entailmentsof the pictorial theory By contrast with these implications of the quasi- in a different manner,our results can be seen as further illuminating the precise conditions under whichsuch perceptual, pictorial mediumtheory, the ’tacit knowledge’ account would predict that the re- seeminglycontradictory results can be obtained. interpretation of imagesis difficult becauseit assumes that the mentalrepresentations are very abstract output "Philosophical" &"Strictly Empirical" of ’higher’ cognitive processes, - encodings of However,it is ultimately fruitless for the protagonists conceptualizationsor beliefs and, in this sense, already in the imagery debate to continue counterposing meaningfuland not requiring interpretation, - nor empiricalresults as if these speakfor themselves.The susceptible of easy re-interpretation (Pylyshyn1973, cloudinessof the presentsituation and the intractability 1978). of the debate is symptomaticof fundamentalconceptual problemsrather than straightforwardempirical issues. Chambers& Reisberg Negative Results Finke (1989) passes over such "philosophical" This questionof reinterpreting visual imageshad been problemsand the principal alternative theory (Fodor brought into sharp relief with the workof Chambers 1975) implyingtheir relative unimportanceand saying and Reisberg (1985) who found that subjects were that "this is strictly an empiricalquestion" whichhas uniformlyunable to reverse their mental imagesof the been decidedby the evidencethat "mentalimages can be familiar ambiguousfigures such as the duck/rabbitand reinterpreted" (1989, p.129). Thecited data of Finke, Neckercube. Chambersand Reisbergsee their results as Pinker and Farah undoubtedlyshow that "imagescan be supporting the "philosophical" argumentsfor taking reinterpreted", but this cannot "refute" Chambersand imageryto be moreconceptual and cognitive, in the Reisbergsince it only begs the central question: What sense that they are intrinsically interpreted symbols is the reasonfor the discrepancy? whichdo not ne, ed, and do not easily permit, further TheChambers and Reisberginvestigation derives its interpretation. Since the close parallel, indeed significance from showing the inability of image "equivalence" of imagery with the mechanismsof reconstrual precisely under conditions in which we perception has been one of the central tenets of the wouldhave expected it accordingto the quasi-perceptual pictorialist theory (Podgornyand Shepard1978, Finke account. By demonstrating the possibility of reinterpretationunder entirely different conditions,Finke 1980, Farah 1988), the results of Chambersand Reisbergare surprising and maybe seen as posing a et al. havemerely avoided the crucial issues, whichare, fundamentalchallenge for the pictorial theory. Of moreover,the "philosophical",that is, theoretical, ones course, these negative results provide experimental Finkeet al., have wishedto dismiss(See Slezak 1990, support for the specific claims madeover a decadeago 1991). Thecommon designation of "philosophical" for the by Pylyshyn(1973, 1978)in his critiques of pictorial theories. Despite these earlier skeptical claims, tacit knowledgeposition by its critics is a peculiar, however, in the intervening period there has been thoughtelling, symptomof the difficulties whichhave experimentalevidence of just suchabilities of peopleto plaguedthe debate. Thenotion of Finkeet al. (1989, p.54) that argumentsin the philosophicalliterature on detect novel properties in imaginedscenes. Pinker and Finke (1980) report subjects’ ability to "see" novel imagery"have no relevance"to the empirical questions properties which"can be ’read off’ the display" and reflects someinvidious, if obscure, comparisonbetween whichshould emergefrom images after mental rotation. rival theories whichare, in fact, identical in status as

13 competingexplanations of the evidence. This attitude construedas knowledge"of their visual systems"or of betrays a certain vestige of positivist or behaviorist neurophysiologyas Farahhas taken it (1988, p.314). faith in the unambiguousdeliverances of observational evidence. Misunderstandingon this point has led to Equivalence Thesis. peremptorydismissal of the tacit knowledgeaccount and to a systematicneglect of its explanatoryforce. Wemay note that a symptomof the problemsinherent Thus, for example, Kosslyn (1980, p.30) has in the pictorial approachis evident in Finke’s (1980) described the tacit knowledgeopposition as a "no discussionwhere he speaksof "physicalobjects" instead imagery"account as if it purportedto denythe existence of percepts of physical objects. He describes mental or reality of visual imagery. It is evidently this imagesas "functionally equivalent to physical objects misunderstandingwhich underlies his tactic of adducing or events, with respect to certain types of effects" further chronometricdata as if they couldadd supportto (,p.113). However,it shouldbe evident that one cannot his theorywhereas, in fact, they can only reinforce the makeany meaningful comparisonbetween images and reality of imageryphenomena which are not in dispute. physicalobjects per se, if the latter are taken seriously This approachamounts to legislating pictorialism true and literally as the relevant items of comparison.That by taking the evidenceto somehowuniquely favour the is, Finke is unlikely to be seriously concerned to pictorial theory. This misunderstandingof the tacit comparethe properties of mental representations as knowledge account is manifest in the design of psychologicalentities with those of tables and chairs experiments which have been taken to bear on the (see also Finke1990, p.171). dispute. Thus,Kosslyn (1983) has tried to deflect tacit By assimilating imageryso closely with vision, knowledgeexplanations of the phenomenaby noting indeedby claimingtheir "equivalence",the pictorialists that "virtually nobodyoutside the field knowsabout the are inevitably committedto predicting closely similar McCullougheffect" and therefore "subjects could not "perceptual" phenomenain imageryto those found in have intentionally or unconsciously produced these perceptionitself. As alreadynoted, the significanceof results withoutusing imagery"(1983, p.81). our non-chronometrictasks as "crucial experiments" First, however,the issue cannot be whether the derives from testing these direct implications of the subjects used imageryas Kosslynsuggests here, since pictorial theory. Specifically, the experimentsare this is not in dispute (see also Finke1990, p.19). concernedto test the ’quasi-perceptual’claims according alreadynoted, the tacit knowledgetheory does not deny to which "the purposes of imagery, in large part, that imageryis used by subjects, but only offers a parallel those of vision" in "the use of recognition different explanation of it. In particular, it is no processes" through which Kosslyn claims "one may argument against the tacit knowledge theory to ’recognize’parts and properties of imagedobjects that demonstrate howcontrol subjects not using imagery had not been previously considered"(1987, p. 149) The give different results fromthose whodo (Kosslyn1981, abilities beingtested are exactlythose whichare claimed p. 238, 142). Althoughthis experimental design is to have been demonstratedby Pinker and Finke (1980) conceivedas havingempirical bearing on the debate, it whoreported subjects’ ability to "see" novelproperties is in principle incapableof doingso. Thecontrol group which could emergefrom images only after they had in all these cases is presumednot to be using imagery been mentally rotated wherenew properties of images and therefore becomesutterly irrelevant to the disputed "can be ’read off’ the display" (1980, p. 262). The question of howthe results are to be explainedin the ability to "re-parse"an imageby using perceptualshape case when subjects are using imagery. Patently, classification procedureshas been claimedby Finke et controllingfor the use of imagerycan onlyestablish the al. (1989), just as Finkeand Slayton (1988) claim empirical importanceof imageryas a variable and does have shown "that people are capable of making not favour the pictorial mediumtheory - unless the two unexpecteddiscoveries in imagery". are implicitly and illegitimately taken to be the same By testing newand quite different "perceptual" question. phenomena,our work shows yet further respects in Furthermore, the undoubted ignorance of the which the alleged "equivalence" betweenimagery and subjectsconcerning such experimental results is entirely perceptionbreaks down.Our experiments can be seen as irrelevant to the possibility of explainingthem through avoidingthe specific criticisms of Finkeet al. (1989), tacit knowledge,since there is nothing in the claim and our data provide not merelya confirmationof the whichrequires that the relevant beliefs be this kind of negative Chambersand Reisberg results, but give a belief. Onthe contrary, the idea of tacit knowledgeis further elaborationof the conditionsunder which images precisely the idea of beliefs whichare unavailableto canor cannotbe reinterpreted. consciousawareness. Pylyshyn has been quite explicit on this question, noting that "muchof it is not Experiment la: Mental Rotation. introspectableor verballyarticulable" (1981, p. 161). Newstimulus materials have been designedto have two Theappeal to tacit knowledgeis to invokewhat the distinct interpretations whichare highly orientation subject knows implicitly about the phenomenaor specific. Thus,the figures are recognizableas a certain events in the world being imagined,and emphatically object in one orientation, but are interpretable as an not about the psychological experiment as Kosslyn entirely different object whenrotated by 90 degrees. seemsto think. Nor, can the knowledgein question be Thesestimuli are variants of the stimuli used by Rock

14 (1973), and are considerably improved in their shapeshaving any alternative interpretation. This was recognizability. In this respect, the shapes have the to set subjects’ expectationsfor the subsequentstimuli importantfeature that the alternative interpretationsare whichwere, in fact, orientation-dependentfigures. For readily obtainedby rotation underperceptual conditions. eachpresented figure in turn, subjectswere asked if they couldrecognize it, to nameit, and then to memorizeit. After 10 secondsthe stimulus figure wasremoved from viewand the subject wasasked to imaginerotating it by 90 degrees in a clockwisedirection. Whenthe subject confirmedthat the figure wasbeing imaginedin this rotated position, he/she was asked if it could be interpreted as anythingelse fromthis viewpoint.This procedure wasrepeated with each of the orientation- dependentshapes in turn. Figure 1 An important methodologicalproblem arises from the danger that subjects might encode both It is importantthat the task of reinterpretation can interpretations during the original perceptualexposure be readily accomplishedin this wayduring perception since in this case they wouldnot be relying on imagery becausethis makesthe conditionsfor reconstrual under processesto discoverthe alternative interpretation. In imageryconditions as favorable as possible. Thus, for particular, it is evidentthat the problemmay arise when example, whensubjects are shownfigure 1 in one subjects are set the imaginalrotation task morethan orientation, it is immediately recognized as the once, since after the first one they are no longernaive duckling; then uponrotating the figure by 90 degrees, concerningthe possibility of a secondconstrual of the subjects immediatelynotice (with frequent expressions figure. Onany subsequenttest, they mayUnavoidably of surpriseand deligh0the alternativeinterpretation, the seek a secondinterpretation during the initial perceptual rabbit. exposure. This possibility can be minimized by restricting exposure time as Chambersand Reisberg have done to a duration which is long enough to establish the image and yet not long enoughto seek alternative interpretations. However,this is a risky proceduresince subjects whoare awareof a possible reconstrual are frequently able to notice both interpretations almostimmediately and simultaneously. Results. At first glance, the experimental results Figure 2 appear somewhat equivocal on the question of reinterpretation since subjects weregenerally able to The direct expectation of the pictorial medium reconstrue in imaginationabout one third (35%)of the theoryis that the sameeffect shouldbe obtainableunder figures they werepresented. Evenon these data it is imageryconditions. That is, if subjects are shownthe clear that reinterpretation of the rotated imagewas figures in only one orientation, it wouldbe expected difficult to perform, even if not always impossible. that they could rotate their image and discover the However,these results across multiple presentations alternative construal by inspection from their rotated take on a greater significance whenthe order of image. Of course, the tacit knowledgealternative presentationis taken into account:It is mostsignificant account takes imagesto be abstract, intrinsically that no subject wasable to reconstruethefirst stimulus interpreted conceptualizationsand wouldpredict that presented, whichis, of course, the only one for which such reinterpretation wouldbe difficult or impossible they did not knowin advancethat there might be an for subjects to performin this wayon their rotated alternative interpretation. This striking relevanceof images. stimulus order supports our conjecture concerningthe effect of loss of naivete regardingthe task. Moreover, stimuli werepresented in order of decreasingsuitability as confirmedby later experiments,and this supportsour explanation for the slight improvementin subjects’ successrate. Wecontrol for these confoundingfactors in the follow up experiments, but even without making suchallowances the meansuccess rate overall wasstill only 35 per cent. Typicalof the predicteddifficulty was Figure 3 the reaction of subjects whenpressed to interpret their MethodWithout marking any distinction amongthe rotated imageof the duckling: just as one wouldexpect stimulus figures, subjects were first shownseveral on a tacit knowledgeaccount according to which the distractors in the form of silhouettes of easily imageis intrinsically boundto its interpretation, many recognizableanimals such as an elephant, ostrich and subjects would volunteer the response that it is a marlin, none of which were orientation-dependent "ducklingon its back"!

15 The significance of ournegative results derives from the fact that the mentalrotation and reinterpretation are Expt. lb. Practice & Perception Effects not only explicitly predicted by pictorial theorists, but The ordering effect in the foregoing data in which involve precisely the mental transformations which subjects showeda slight improvementfrom their initial have been classically taken as well established. Of failure could be due to practice in the task rather than to additional importanceis the fact that our task is readily perceptual confounding as we had suggested. We performedunder perceptual conditions, thereby entitling controlled for this possibility by giving each subject us to expect it in imagery as well according to the prior practice with image rotation using Cooper’s pictorial account. Further favouring reconstrual is the (1975) random polygon experiment. fact that our figures are considerably simpler than the At the same time, in order to preserve subjects’ representations of blocks stacked in three dimensions naivete on all the stimulus figures, we altered the employed by Shepard (1971) for which the mental previous instructions so that the imagery task would rotation has been claimed, and our own shapes are not be knownuntil all the figures had been viewed and geometrically no more complex that those of Cooper memorizedin their initial orientation. Onceall stimuli (1975) for which complexity was specifically found not had been memorizedin this way subjects were prompted to be a factor in the claimedease of rotation. by the brief flash of a figure on the screen, and asked if Even if the Finke et al. (1989) response they were able to recall the shape clearly. Only then Chambersand Reisberg (1985) is not problematic in the were subjects asked to rotate the shape 90 degrees waysI have suggested (Slezak 1991), their "refutation" clockwise and asked whether they were able to find an is specific to the ambiguous stimuli and, therefore, alternative interpretation. irrelevant to our entirely different imagery tasks. A pattern of such failures on diverse perceptual phenomena Results: Significantly, prior practice with rotation of would leave only ad hoc ways of avoiding their images on the Cooper random polygons had no effect significance for the pictorial theory of imagery. on subjects’ performance and this possibility could, Therefore, as a follow-up, we have devised additional therefore, be eliminated as a possible explanation for the experiments which attempt to reconstruct yet other ordering effect in the preceding data. Indeed, despite perceptual phenomenain imagery. practice in rotating images, there was a dramatic drop in the success rate as a consequenceof the newstrategy to Experiment 2. Figure-Ground Reversal avoid perceptual confounding. Our data show only 8 successful reconstruals in 100 trials, and these were The shapes illustrated in the left half of figure 4 are almost entirely confined to two of the figures whose such as to encourage perceptual organization into shapes were said by subjects in debriefmgto be a "give- several black objects which may, however, be reversed away"due to certain telling clues. to becomethe groundand thereby reveal the letters "EI". Since the reversal in this form is somewhatdifficult to Experiment lc. Image quality. achieve in perceptionitself, the effect can be elicited by It could be argued that under the new conditions for asking subjects to being the horizontal lines together to avoiding perceptual reconstrual, the high failure rate was touch the shapes as in the right hand figures, clearly now due to poor, inaccurate or otherwise degraded revealingthe letters. images. In order to clarify the possible role of this factor, we altered the conditions in such a way as to maximizethe accuracy of encoded shapes in memory. Subjects were tested on the imagery rotation task nowonly after being permitted a very long (3 minute) visual presentation of one of the stimuli (following the usual distractors). During this extended viewing time, subjects were encouraged to rememberdetails of the single figure as accurately as possible. Whenthe stimulus was removed, subjects were asked to draw it from memoryin order to have some evidence of the Figure 4 accuracy of the image. Subjects were then asked to rotate and reinterpret it in imagination.In addition, D.F. Despite the ease of the imagery task, not a single Marks (1973) VVIQ(Vividness of Visual Imagery subject in twenty trials was able to reconstrue their Questionnaire) was administered for further evidence of imageto reveal the alternative construal as letters. This image accuracy. was despite the fact that subjects’ subsequent drawings of their memorizedimage were highly accurate. Results and Conclusion.: Despite high scores on the VVIQaveraging 2.5, there were still only 2 Expt. 3. Kanizsa Illusory Contours successful reconstruals in 23 trials. Aboveall, the accuracy of the drawings nowprovided fu’m grounds for Stimuli of the sort illustrated in figure 5 were designed supposing that degraded image quality is not a likely to produce the familiar illusory contours, but were, of reason for the failure of imagereconstrual. course, not presented to subjects in this form, since the

16 effect wouldthen be created in perception.In order to Farah, M.J. 1988. Is Visual ImageryReally Visual? test the parallel with imagery, circumstancesmust be Overlooked Evidence From Neuropsychology. contrivedwhich generate the figure only in imagination Psychological Review 95:307-317. and, accordingly,the entire figure wasnot presentedat Finke, R.A. 1980. Levels of Equivalencein Imagery once. Instead, the apical black shapeswere designed to and Perception. PsychologicalReview 87:113-132. havegood gestalt propertiesas figures andarc suchas to Finke, R.A. 1989. Principles of Mental Imagery. discourageany inferences about other shapes of which Cambridge,Mass.: BradfordBooks/MIT Press. they might be a part. Theseshapes were presentedone Finke, R.A. 1990. Creative Imagery. NewJersey: LEA. at a time for 30 secondsat their respectivepositions and Finke, R.A., Pinker, S. and Farah, M.J. 1989. then removedfrom view. Havingseen themonly one at Reinterpreting Visual Patterns in MentalImagery. a time, subjects were instructed to imaginethem all Cognitive Science 13:51-78. together and werethen askedwhether they wereable to Finke, R.A., & Slayton, K. 1988. Explorations of detect anyother emergentshape, figure or object in their Creative Visual Synthesis in Mental Imagery. reconstructedimage. Memoryand Cognition. 16:252-257. Kosslyn, S.M. 1980. Image and Mind. Cambridge, Mass.:Harvard University Press. Kosslyn,S.M. 1981. The Mediumand the Messagein Mental Imagery: A Theory. Psychological Review 88:46-66. Kosslyn, S.M. 1983. Ghosts in the Mind’s Machine. NewYork: W.W.Norton & Company. Kosslyn, S.M. 1987. Seeing and Imagining in the Cerebral Hemispheres: A Computational Approach. Psychological Review 94:148-175. Kosslyn, S.M. 1988. Imageryin Learning. In Michael Figure 5 S. Gazzanigaed. Perspectives in MemoryResearch, Cambridge,Mass.: Bradford/MITPress. Results: Despite research on "creative mental Larkin, J. and Simon, H 1987. Whya Diagram is synthesis" whichsuggests that peoplecan use imagery (Sometimes) WorthTen ThousandWords. Cognitive to mentallyassemble the separatelypresented parts of a Science11: 65-99. pattern (Finke 1990, p.21), only one subject out Pinker, S. and Finke, R. 1980. Emergent two- thirty trials reported seeing a geometrical shape, DimensionalPatterns in Images Rotated in Depth, correctly identifying the emergentwhite figure. Again, Journal of Experimental Psychology: Human this overwhelmingdifficulty with the task wasdespite Perceptionand Performance6(2):244-264. the fact that subjects’ drawingswere highly accurate and Podgorny, P., and Shepard, R.N. 1978. Functional they werefrequently able to notice the emergentshape Representations Commonto Visual Perception and fromtheir owndrawing. Imagination. Journal of Experimental Psychology: HumanPerception and Performance4:21-35. Pylyshyn, Z. 1973. What the Mind’s Eye Tells the Conclusion. Mind’s Brain: A Critique of Mental Imagery. Notwithstandingthe claim by Finke et al. (1989) PsychologicalBulletin. 80(1): 1-24. have "refuted" Chambersand Reisberg (1985), we have Pylyshyn,Z. 1978. Imageryand artificial intelligence. shownthat imagereconstrual is generally difficult or In MinnesotaStudies in the Philosophyof Science, C. impossible to performunder conditions in whichone WadeSavage ed. Vol. IX. Minnesota: University of wouldhave expected it accordingto the pictorial theory. MinnesotaPress. Onthe other hand,however, these results are precisely Pylyshyn, Z. 1981. The Imagery Debate: Analogue as one wouldexpect on the tacit knowledgeaccount Mediaversus Tacit Knowledge.Psychological Review, according to which imagery is highly abstract and 88: 16-45; reprinted in N. Block(ed.) 1981. Imagery. cognitive, and doesnot involveany internal, surrogate, Cambridge,Mass.: Bradford/MITPress. diagrammatic’objects’ to be apprehendedby the visual Reisberg, D. and Chambers,D. 1991. Neither Pictures system. Nor Propositions: WhatCan WeLearn from a Mental References Image?Canadian Journal of Psychology,45 (3): 336- Anderson, J.R. 1978. Arguments Concerning 352. Rock, I. 1973. Orientation and Form. NewYork: Representations for Mental Imagery. Psychological Academic, Review 85:249-277. Shepard,R.N., and Metzler, J. 1971. Mentalrotation of Chambers, D. and Reisberg, D. 1985. Can Mental three-dimensionalobjects. Science171:701-703. Images Be Ambiguous?Journal of Experimental Slezak, P. 1990. Re-Interpreting Images. Analysis Psychology: HumanPerception and Performance 50(4):231-243. 11:317-28. Slezak, P. 1991. CanImages Be Rotatedand Inspected? Cooper, L.A. 1975. Mental Rotation of RandomTwo- Proceedings of 13th Annual Conference of the DimensionalShapes. Cognitive Psychology7:20-43. Cognitive Science Society, 55-60. NewJersey: LEA.