SPACE PERC

Organisatew: 6. JOHANSSON ( Uppsala University,

6. OLDFIELD (ht. of Exp. Psych., oxford, fig

Rappsrteurs: W. M~ZGER, I. GIBSON, G. JOHANSON, 9.

Discutants Wit&: I. KOWR, M. D. VERNON

WAS’ IST WIRKLICH NEU IN DEIt THEORIE DES

EN SEHENS?

VON

WOLFGANG ETZGER

(Habn’cktshtihe 26, Miinster / W’estpblen)

Uber das binokulare Tiefensehen haben wir in letzter Zeit gelernt: 1) aus den Berechnungen und Beobachtungen von Ames:

die ausserordentliche Feinheit und Bestimmtheit der binokularen Effekte;

2) aus den Untersuchungen von Heinz Werner: a) dass blosse Annaherung - such ohne usion - sehon eindeutige

Tiefenwirkung hat, b) den Trennungs’verzug binokular vereinigter Konturen. c) dile Zunahme der Attraktion bei abnehmendem Abstalrid,

us b) und c) folgt die Verstiirkung der Tiefenwirkung It& bewegtem

3) aus der grossen Ar’beit Linschotcn’s unter vielem anderen :

a) &ss zur vollen Fusion ausser der angenaherten formalen such

258

nture~ sich nur an einander lagern,

imensionaler Gebilde. ruitidlagen der Unver-

ss er sick iiber die ernsten Zweifel, die rden, nicht wundert.

figurationen getrennt besprechen. ar sein, der frisch und ohne vie1

ungen herzustellen suchte und

s man aueh das, was wir sonst als Reizkon$gurationen und ak

Dagegen bin ich nach wie vor der einung, dass die offenbare Empfindlich- keit des reti;?alen osaiks fur solche Konfigurationen und ihre Trans-

i der Untersuchung der zuerst von usatti und Renvall mit- geteilten Erscheinungen, ftir die Wallach inzwischen den Namen ,,kinetischer Tiefeneflekt“ eingeftihrt hat, bin ich seinerzeit einen Weg gegangen, der dem Weg Gibsons genau entgegengesetzt war: ich versuchte damals, zuniichst nicht optimaie sondern minimale Bedingungen herzustellen, und fragte : welcbe Bedin en miissen erftillt sein, damit geracie eben eine Ticfenwirkung auftritt. ies war der Sinn des Experimentrierens mit den Schatten gleicher und paralleler St&be auf einer Drehscheihe in paraheler Brojektion. Was hier in letzter it experimentiert wurde, brachte nicht vie1 Neues, Wallach’s Schlussfolgerung, dass im Schattenversuch a) Lsngen- linderungen, b) Winkeltinderungen Mtig und ausreic end seien, um &here Tiefenwirkungen zu erzielen, war schon 1934 tiberholt : die Schatten van 3 Stsben in der Anordnung eines un lmtissigen Dreiecks geben ohne diese Bedingungen einen ebenso starken Effekt. Die Schatten eines um seine kurze Diagonale rotierenden Rhomhs erfiillen beide Bedingungen, ergeben aber nut einen schwachen, wenig zwingenden und wenig bestgndigen, bei vielen Vpn. iiberhaupt keinen Tiefeneffekt. Seine Meinung, dass der

is durch eine e deckenden kJeinen Achsen, so wird ctie

darin, dass jerie nur eine geneigte R&z

zeichnet und seine bestindigere rlumliche Wir&~ tonung der Starrheit der gesehenen Gebiide ist wichtig; &s

Verhalten des Rhombenschattens weist aber - wie ich schon 1934

Tatsache verdankt, dass sie das einfachste unter den herrschende

Rhombenfigur der Fall ist, verliert die Starrheit ihre Vorherrsch tie es nach der Mgnanztheorie Wertheimer’s zu erwarten ist. I&se ist also trortz Ho&berg’s Uberiegungen und Versuchen bisher nicht ersetzt.

Bei den Wirkungen ruhender Konfigurationen kommt nichts den Gib son’schen Texturgradienten nahe, iiber die in let&r Zeit hum diskutiert wugde.

Umsomehr Arbeiten beschgftigen sich mit den zuniiichst von Kopfermann untersuchten dreidimensionalen Wirkungen einfacher geometrischer Strich- zeichnungen.

Hier erfolgt fast gleichzeitig der Versuch von Eb und der von Ishii, das Ganze - unter Verzicht auf die Mgna s Tiefentendenzen ein;reher Strichgabelunen oder schiefer Winkel zu verstehen und die Gesamttiefenwirkung als die Summe dieser Einzel-Wirkungen zu etkl%rcn. Diese Erkhirung ist schon 1930 von Kopfermann selbst nwide ihre Argumente werden in neueren Arbeiten einfach ilbe n~m.hch bei icher Zahl von schiefen Winkeln und Ga Cesamtgebil schon zweidimensional genilgmd symmetrisch sind, dcr TieftnefKekt ausbleibt.

.A uch der Versuch Wallach’s, den Kopfermann-Effekt als Machwirkung iefeneffekts zu e&h&en, scheint mir verfehlt. Nach eigenen

neuen Uatersuchungen ist die prozentuab Vermehrung der ,,Tiefen“-

stiitzt, nach ents~r~hend~n usatti hatte sie scho in den zwanziger

em bleibt die bevorzugfe iefenwirkung zwei- m~tri~her ~~bi~de erh Jten und durch die wegungs-

y means of' such optical instruments as the spectroscope, the physicist can obtain i~f~~ti~~~ from the li t which enter:: the instrument. For example, he can get information about the temperature of the ligh from the distribution of wavelen s, or spectrum, of the light. discover the chemical compositi f the source from the dark lines of this s trum, and the approach 3r recession of the source from the Doppler efbt on these lines. He can know some of the prop&ties of the transmitti medium by measuring the polarization of the light, or its diflraction. Avd so on.

But this is radiant Ii t. The perceptual psychologist is primarily interested see things, that is, light which illuminates a physical

ich is scattered by diffuse reflection from solid surfaces. ht in a transparent medium such as air. it trave

back and forth in all directions. Since the travel is strictly rectilinear, a projection of the different reflectances of the solid surfaces is obtainable at an int in the air.

is considered in this way, as converging to a point instead om a source, it carries information of a different kind from

that obtained by the physicist--information about the environment. This is the information which the eyes of animals are capable of registering. The analysis of this information has been neglected, and the very possibfity of it has been denied. It requires a science of what might called “‘ecological

optics” as distinguished from physical optics. Only the beginnings of such

a discipline can, as yet, be suggested. The focussable light projected to any station point in the air can be

termed an optical UWUJL For any environment except an empty blue sky,

262 18

this array will have a structure or compositions in rent dire&i0 textures, contours, or forms in the array. ometrical differences remains to be settled.

r sensations, or the patchwork of eel is not relevant, although it may be su

Some information about the space of the environment is carried by the structure or pattern of an optieal array to a single station point. try of perspective can be used to analyse this information, information about space is carried by the chge of pattern of an opticat array to a moving station point. The geometry of continuous perspective transformations can probably be applied to this information. Similarly, for animals like man whose eyes point forward, further supplementary information about space is carried by the dzJi?rence or disparity of pattern between the optical arrays to slightly d&krent station points. the geometry of transformations can be applied. It is reasonable to suppose that the ocular system is sensitive to the pattern, the change of pattern, ;mnd the &parity of pattern of fwussable light.

As an example of a simple but important type of environmental space, consider the edge of a horizontal surface, in short a cliff. An animal citn &z;~cend a short drop-off but will fall through a long drop-ofTand be injured. it should therefore discriminate between the two and avoid the lattc:r. According ts the experiment of Walk, Gibson and Tighe (Science, 1957, 126, 80-81) rats do so (and incidentally do so even if they have been reared in complete darkness). The optical array to which the rat must respond in this experiment is illustrated, and the geometrical variables to which. it responds are pointed out.

Considering both optics and ecology, a number of variables of the arl’sy of tight are analysed for what they indicate about the environment (optic4 texture, contour, closed contour, density of texture” together with ratios and gradients of these variables). This analysis of pattern is contrasted wi,th the laws of visual organization proposed by Gestalt theory. In particu- lar, experimental evidence will be reported to show that certain trans- fchrmations of pattern in an artificial optical array induce perceptions of space which would in fact be correct if they occurred in the natural optical array resulting from the ordinary motions of objects.

The problem of blow the ocular systems of ditrerent animals register the potential information in the flux of light surrounding them is quite a cii&zrent problem from the above. For this we must understand the ocular responses of fixation, compensatory movement, saccadic movement,

PE ON BE ~'~PAC~ 263

~~~~en~ of two eyes, in addition law adjustments that yield retinal images.

” do not, it must

re also taken into account, urate, and vast in amount.

RI

BY

GUNNAR JOHANSSON

(Uppsaia University, Sweden)

A discussion of some mathematical principles concerning the perception of relative motion and of their ssible function as determinants of static perceptual space.

In the past, too much of the theorizing about space Gerception has been based on an assumption of a static pattern of excitation.

Recent investigations have very definitely indicated that continuously changing retinal excitation is not only a possible, but apparently alsti a necessary condition for perception. The paper deals with experiments conducted by Ratliff, Ri s and other investigators, where a mirror attached to a contact lens is used for the absolute fixation of an image on the retina, as well as wib:+ Hochberg’s “Ganzfeld” experiments.

The purpose cf the present paper is to demonstrate how, in its function, event perception reveals astonis’hingly strict mathematical principles, and to show that, for many problems in space perception, change perception and event perception are more natural starting-points than those enlployed in the classical approach.

To a very great extent, perception of real motion is an unexplored field- The writer considers that it is a very fertile field when it is a question of solving more general problems of space perception. The most important contributions towards the understanding of the perception of “real” motion

and his ccbworkers at Corn nivenity have aSso cellent studies.

The author is also interested in motion ally to n a clearer insight into the mechanisms o of relative motion. In this end of the investigators mentioned above.

Studies in relative motion reveal that our vision functions not as a cinematographic camera, but tather in close analogy with a mathematical computer able to provide a differential and inte 1 calculus if the stimulus correlates to motion vectors in space. Thus, we t to the physial motion tracks in an image, but to their abstract ~e~ti~~l components or resullants.

Our visual apparatus “takes out” and calculates, alon one common time atxis, equal motion vectors in Ihe proximal stimulus paWms. These ~mmon vectors (motions simultaneously equal in direction and :magnitude) combine perceptually and form a unit of emotion. Such a uuit acts as a frame cdf reference for the clivergent motion vectors which remain as a residual” Divergent vectors, qwhich are equal respectively, comMne to form another unit, etc.

During the symposrum a Elm was shown to illustrate such a perceptual motion analysis of proxim stimulus in vectors. This al@ysis may be termed the principle OF the “common motion state”.

oreover, it is rceg’arded as demonstrating a neral perceptual principle, the attempt to attain the highest possible of simplicity together

with the least possible degree of change. At the Cornell perception, when discussing this perceptual principle, it was t minimurn principle”‘.

There was also a demonstration of the perceptual tendency to avoid, if possible, changes in the change, i.e. expressed in mathematical terms: derivates of higher order. This may also be considered as an effort to admit the least degree of &ange. From this point of view it provi an “explana- tion” of a depth efiixt often found in experiments with -dimensional harmonic motion.

I have here discussed a perceptual analysis of a pattern of motion in the total proximal stimulus, as if this pattern was present also in excitation. By greatly s~irnp~yin~~ actual conditions, the excitation may, however,

consideredi to cons& of two components : (a) the pattern formed by the array of light in proximal stimulation, and (b) the eye movements.

al stimulus, as such, is he retina, but it may

the divergent component

ulus pattern, which is a correlate of strict mathematical component

y of our visual apparatus to ur studies in relative motion

much in support of the hypothesis that the type of analysis discussed here abstracts the invariance from change, and from

retinal e~cita~on, id static space. discussed which indicate that proprioception tes directly to visual motion perception.

SPATIAL. ~EP~ESE~TATI~~

BY

JULIAN E. HOCHBERG

(Cornell University, Ithaca, N. Y.)

The study of space perception had roots in two early sources. The first, was tbe practical problem of pictorial depth representation, typified by Da Vinci’s analysis of the portrayal of space--the visual depth cues derive largely from such investi tions. The second source was Berkeley’s episte- ~nclogical Query concerni how we can know the “external” world: Berkeley pointed out tha depth cues are not necessariiy correlated with spatial distince, and maintained in consequence that all of our ideas about space must rest ultimately upon the non-visual sensations of accomodation,

rice, etc, to which the several cues have merely become associated through habit. Most of our “major theoretical issues” (empiricism, associ- ationism, atomism) derive from this secosnd source. eanwhile, although

the pictorial depth cues are continually cited, their basic psychophysical investigations hbve been unjustifiably neglected.

Three theoretical obstacles have been raised to the psychophysical study of depth representation (ati three viewpoints appear to assume that stimula- tion and response must be in perfect correspondence for such study to be of value, revealing their epistemological preoccupation) :

266

1) Training, motives, etc., may (in certain tion--“ therefore’“, perception is idiosy

2) ictured depth relationships czn be reve ---“therefore”, depth is phenomenally ous and “spool”.

3) esponses to two combined stimuli are fr~u~~y quite those each stimulus ekits sepBra~~y-“therefop’, actions rule, and we must turn to the study of the iso logical substrate to bring order from chaos.

Let us now consider each of these three theoretical positions in the light of empirical evidence. 1) Training, motives, etc., do affect certain onses to ozrtain pictorial

stimuli; however, to the extent that such e hWt! held impcrtailt in pictorial perception, just so much have they also n attributed to

presumably stable *judgments--brightness* color, size., etc.-which are traditionally lawful and useful psychophysical dimensions. In fact, very low intersubject response variabilities can be found with pictorial stimuli (shdcs 1 2 show group distributions of ju on pictured linear rspective and relative size). After arithmetic symbols are quite systematic even tho learned.

2) Bhenomeaa9ly, there is frequently an enormous difference between pictured and real space, but the varieties of spatial distance, volume, etc.) are complex zind unexplo in any event, pictorial space is not thereby rendered either illicit or uninvestigable. More objectively, “ambiguity” means that the same stimulus elicits different perceptual responses at different times : a) Such ambiguir;y is not restricted to pictures.

ant.ennae suddenly reverse apparent shape and binocular vision, like isometric drawings. In monocular experi- ments with moving targets, slant judgment errors closely paralle:led

ose obtained with stationary perspective drawin s of the targets. Gibson is right, that real spatial motion theoretically can yield more rendundant information ,811 pictures, but the extent to which the information is used the organism is an empirical question. Empirically, the ambi y of pictorial space can be very low, just as that of real space can be high.

b) Moreover, even high ambiguity may be lawful, in two senses: firstly, the degree of ambiguity may be predictable from stinlulus characteristics (slide 3). Secondly, the frequency of appearance

N BE L'EBPACR

ed to measured stimulus little residual variance. taltists made a promis-

owever, after twenty : psychophysi~lly, “

“cortical organization” es may eventually

y we need not wait: or by any other ationships which

do not exist; if such ~y~~o~hysi~~ functions do exist, we must find them (and may u fore the “true” underlying pro- cesses can t. Since psychophysical regularities do exist in pictorial representation (slides 5, 6 7), all the raw materials necessary to derive usable Iaws are obtainable now. After all, the Gestalt “(r,w$’ of organization were discovered without reference to auy physiological processes at all; even if we assume that these “laws” would survive more objective definition and test than they have yet received, they would themselves provide proof that we can state lawful relations without invoking internal intervening variables.

b) No matter how complex the underlying Gestalt relationship, it is mathematically guaranteed that we can always substitute a simple polynomial function which will predict our data with vanishingly little more error than the “true” function. (Slide 8 compares the spatial responses to a large number of isometric perspective figures, and the theoretical values obtained by inserting measured stimulus chara.cteristics in one predictive psychophysical function). The predictive function used here, as an example, is (in words) that “with stimuli to which more than one response is likely to occur, the relative probability of obtaining a given response, is, other

equal, inversely proportional to the amount of information required to specify the object of the response”. This merely attempts a measurable version of “simplicity”, “homogenleity”, etc.-a quantitative extension of Kopfermann”s brilliant work.

The present task of perception psychologists is to predict subjteets’ perceptual responses, from the characteristics of the stimulus field. Gibson has shown the possibility of such prediction for the relaiivel,y determinate world of three-dimensional motion; I have tried to show that the task is equally hopeful in the relatively less determinate sphere of’ spatial rePreSen- tation.

Iv0 Kornra : First sf all, we all agree that the so-4 something trb 153 with our perceptions. This excellent solved all probl.ems of perception already hundred yc~~ possible to find a simple function of correspondence be sthdus responses. We further agree that, since this was not the case, it would not. be right to draw the conclusion that “since there is no simple function, there is no fun&ion at all”. However, the problem to be solved is what ought xo be done if no simple correspondence is to be found and if-as it is the: cxvz-we do not believe in telepathfl This is the basic problem of percepti .

My replay--;rnd also that of others-is: It is n#.xessary to take the greatest possib’;e number of conditions of the entire strinulus situation into account. It might be that not single stimuli, but complexes of stimuli (in the sense of highly inetarcorrelated coincidences of stimuli) are the real stimulus for perception. In this connection a technical detice *) is demonstrated, which responds not only to stimuli, but also to stimsllus patterns (independent of individual stimuli!). The relation to the oints of view expressed by Gibson, Johansson and Hochberg is obvious.

From this point of view only the reproach of being “luxurious” can be raised against the Gestalt-concept of space perception (accordin ger). The description of the phenomen is in both cases the same, but the explanation is more parsimonious if “intervening variables’* (in the sense of autochtonous tendencies of organization) are renounced until, within the stimulus situation itself, constellations are found which are more srdde

ban the individluul stimuli, and as long as it is found that just these invariant constellations are the stimulus for spaci; perception.

M. D. VERNON: The point which to me is most striking in the phenomena discussed in thi,s symposium is their lability. They are perceived by a certain proportion of ithe observers for a part of the: time which they are viewed.

ut they are variable, and are never perceived uniformly by everyone; whereas in the majority of cases the objects and spatial surroundings which we view in everyday Me can be perceived at any time by anyone who is looking at them. Therefore I cannot help concluding that the laws and principles which have beeu hypothesized by the speakers in the symposium from the phenomexx4 discussed cannot be extended beyond the narrowly

--

* The descript.ion was published in Sru&um Gemruik, vol. 10, fwz. 6, 1957, p. 341.

e some measure of the

of revealing relief and texture. I hope to publish some account of the development of these ideas in due course.