Neuroscience Letters 281 (2000) 119±122 www.elsevier.com/locate/neulet

Neon colour spreading in three-dimensional illusory objects in humans

Marja Liinasuoa,*, Ilpo Kojob, Jukka HaÈ kkinenc, Jyrki Rovamod

aInstitute of Biomedicine, Department of Physiology, P.O. Box 9, 00014 University of Helsinki, Helsinki, Finland bThe Finnish Institute of Occupational Health, Brain Work Laboratory, Topeliuksenkatu 41 a A, 00250 Helsinki, Finland cDepartment of Psychology, General Psychology Division, P.O. Box 13, 00014 University of Helsinki, Helsinki, Finland dDepartment of Optometry and Vision Sciences, University of Wales, College of Cardiff, P.O. Box 905, Cardiff CF1 3XF, UK Received 11 October 1999; received in revised form 7 January 2000; accepted 10 January 2000

Abstract We studied whether neon spreading can be induced within three-dimensional illusory triangles. Kanizsa triangles were induced by black pacman disks consisting of red sectors with curved sides. Viewing our stimuli monocularly produced two-dimensional illusory contours and surfaces as well as neon spreading in each ®gure. Triangles appeared concave or convex under stereoscopical viewing. Neon colour spreading was induced within illusory ®gures bending in three-dimensional space, suggesting that neural contour completion and surface ®lling-in interact across depth. Surpris- ingly, neon spreading was induced above the intervening surface even when the inducers were below the surface. Neon colour and illusory con®guration were preserved behind the intervening surface only when it appeared transparent. q 2000 Elsevier Science Ireland Ltd. All rights reserved.

Keywords: Vision; Perception; Illusion; Three-dimensionality; Colour; Neon spreading

Illusory ®gures with one to three dimensions are created zontal and vertical crossing lines so that some line elements by the neural processes of visual area V2 [15] between some are replaced by different colour or luminance, the surround- physically existing objects, known as inducers. A related ings of the line with deviant colour or luminance appears to phenomenon is neon colour spreading [14], the apparent be ®lled with this colour or brightness [13]. In illusory spreading of colour or luminance (cf. ®lling-in within V1 ®gures, deviant colour or brightness seems to produce a [7]) to its neighbouring area without this physical quality. uniform spreading within the illusory ®gure [16]. Then, Classically, neon colour within illusory ®gures is produced the illusory area looks transparent and colour-tinted or by using a colour or luminance different from the back- different in brightness. ground and inducers. This makes the illusory ®gure to In early vision, illusory contour [15], surface, and neon appear as colored, transparent and often glowing [1]. spreading formation is prevented by a nearer surface [3]. An example of a one-dimensional illusory ®gure is an This has been assumed to re¯ect early neural processes illusory contour created between two abutting gratings [4]. Based on this, we wanted to study whether neon spread- with a phase shift [11]. Two-dimensional illusory ®gures ing in stationary images extends across several depth planes have contours and surfaces in one depth plane. For example, when inducers and the induced contours/surfaces have three pacmen (disks with a sector removed) forming the different depth signs. Hence, we used three-dimensional apices of a triangle usually elicit the perception of an illu- Kanizsa-type triangles, somewhat similar to the curved sory Kanizsa [6] triangle. Perceptions of thin Kanizsa Kanizsa-type squares [2], but produced by stereopairs of squares bending in 3-D space have been produced with black disks with red sectors forming the apices of a triangle stereo pairs [2]. on a white surface and bending in depth, thus, making illu- Neon spreading in stationary stimuli seems to be two- sory triangles to appear convex or concave towards the dimensional [1]. In a line drawing consisting of black hori- observer. We recruited 18 ®rst-year students of psychology without * Corresponding author. Tel.: 1358-50-307-5955; fax: 1358-9- experience in perception studies. All had normal vision, 191-8681. with or without correction, and received course credit for E-mail address: marja.liinasuo@helsinki.® (M. Liinasuo)

0304-3940/00/$ - see front matter q 2000 Elsevier Science Ireland Ltd. All rights reserved. PII: S0304-3940(00)00818-1 120 M. Liinasuo et al. / Neuroscience Letters 281 (2000) 119±122 their participation. It was checked that participating students sectors were always on the same depth level relative to each were able to perceive (i) stereo ®gures by presenting red and other. Only the three-dimensional curving and location of green stereo pairs, (ii) illusory ®gures by presenting a the Kanizsa triangle, relative to the surround, were varied. Kanizsa triangle with three black pacmen drawn on a In the ®rst and second stimulus, the tip and the horizontal white paper, and (iii) neon colour spreading by presenting edges of the lower sectors had null disparity with respect to a photo of black pacmen with red sectors usually resulting in the contours of the inducers, and should be perceived to be the perception of neon spreading within a Kanizsa triangle. in the same plane. The sectors of the ®rst stimulus should Five students were eliminated due to failure in perceiving appear convex and above the level of the black inducers neon colour spreading. Thus, 13 students (aged 19±35, mean (Fig. 1a, the two on the right/left ®gures for crossed/ 23 years) served as subjects in the experiments. uncrossed fusion). In the second stimulus, the sectors should The stimulus ®gures were drawn with Illustrator, printed appear concave and further in depth than the black inducers on ®lm with Digital Palette equipment, and presented as (Fig. 1a, the two ®gures on the right/left for uncrossed/ colour prints using a lenticular stereoscope, Stereopticse crossed fusion). (EARTHings Corp., USA) allowing stereo pair observation In the third stimulus, the tip and horizontal edges of the by means of parallel fusion. The device is made of black lower sectors should appear to be further from the observer cardboard and has holes with lenses allowing eyes to focus than the inducers. The rest of the convex sectors should on the plane of the stimulus. Viewing distance was 13.5 cm. appear closer than the inducers and even penetrate the The stimuli were black inducers with white surrounds and white ®eld bearing the inducers (Fig. 1b, the two ®gures red sectors forming the apices of Kanizsa triangles (Fig. 1). on the right/left for crossed/uncrossed fusion). For all stimuli, the inter-centre distance between the pacmen The fourth stimulus was the reverse of the third. The tip was 2 cm (8.58) and their diameter was 1.5 cm (6.48). The tip and horizontal edges of the sectors should appear closer to of the upper sector and the horizontal edges of the lower the observer than the inducers. The rest of the concave

Fig. 1. Stimuli used in the experiments. Although only parallel fusion was used, stimuli for converging fusion are also provided. The instructions are given for parallel fusion; for converging fusion, the opposite applies. Thus, when instructed to fuse the two leftmost ®gures, the convergers should fuse the two rightmost ®gures and vice versa. (a) Fusing the two leftmost ®gures (®rst stimulus) usually produced the perception of a convex illusory triangle with a reddish surface above the contours of the black inducers. Fusing the two rightmost ®gures (second stimulus) resulted in the perception of a concave red triangle. Typically, only its apices were visible through three holes showing a black background, without illusory contours or neon colour spreading. (b) For most observers, fusing the two leftmost ®gures (third stimulus) produced a perception of a convex illusory triangle with neon colour spreading. The apex of the upper sector and the horizontal edges of the lower sectors were further away from the observer than the contours of the black inducers; the rest of the illusory ®gure curved in front of them. The black areas were perceived as holes or funnels. When the two rightmost ®gures were fused (fourth stimulus), most observers perceived a red concave triangle, only its apices appearing to be closer than the contours of the black inducers. The rest of the sectors appeared to be further away from the observer than the white surround that appeared to occlude the central part of a real triangle. M. Liinasuo et al. / Neuroscience Letters 281 (2000) 119±122 121 sectors should reach into the distance, penetrating the white subjects out of 13. When the above mentioned parts of the ®eld so that they are further from the observer than the sectors were closer to the observer than the rest of the inducers (Fig. 1b, the two ®gures on the right/left for con®guration, a concave, partly-occluded triangle was uncrossed/crossed fusion). usually reported, and the neon effect was reported less The subjects observed each stereo pair using Stereo- frequently, by only two (second stimulus, x2 1†ˆ121, pticse and reported their impressions aloud. After the P , 0:001) and one (fourth stimulus, x2 1†ˆ144, subject gave a spontaneous response, the experimenter P , 0:001) subjects. enquired, if necessary: (1) whether the subject saw the illu- Although illusory contours existed without neon colour in sory triangle: (2) if yes, whether the subject saw contours one case (subject S.R. with the ®rst stimulus), no one saw around the triangle; (3) about the quality of the illusory neon effects without illusory contours. The other individual surface, and (4) about its three-dimensional properties, differences in the perceptions seemed to depend on the that is, were some parts nearer or further away from the perceived location of the white surface with respect to the observer than others. red sectors and whether or not the surface appeared trans- Twelve subjects out of 13 perceived neon colour spread- parent to any degree. Subjects S.S. (the ®rst stimulus) and ing in the ®rst stimulus. Eleven of these perceived a reddish S.R. and S.S. (the third stimulus) perceived the relative convex illusory triangle with its apices on the pacmen but depth of the illusory or partly occluded triangle differently otherwise in front of them and `above' the white surround. from the other subjects. This could be due to a subtle de®cit One subject (S.S.) perceived the illusion in the same way but in stereo vision. behind the white surface that was transparent to some We found that three-dimensional neon spreading can also degree. Subject (S.R.), who did not perceive neon colour be produced within steady illusory ®gures curving in depth, spreading, described a glossy uncoloured transparent trian- extending previous ®ndings [1] where neon effects were gle curving above the white surface. obtained in three-dimensional settings but with ¯at shapes. In the second stimulus eleven subjects did not perceive Junctions seem to be important in creating neon colour neon colour spreading but saw the pacmen as holes through spreading in ¯at frontoparallel illusory ®gures [17]. Our which the three apices of a red concave triangle were visi- experiments show that these junctions do not have to consist ble; the other parts of the triangle were behind the white of planar curves but neon effect can also spread from surface. Thus, the whole triangle appeared to be deeper in elements curved in depth. depth than the white surface. Two subjects (T.T. and M.K.) Illusory contours, surfaces and neon spreading have been perceived an illusory triangle with neon colour spreading shown to be prevented by a nearer sharp-edged object [3]. behind the white surface bearing the contours of the holes. This has been assumed to re¯ect early neural processes [4]. The white surface appeared suf®ciently transparent to reveal However, our experiments with stimulus three showed that the red illusory surface, which was not transparent enough spreading can be induced above the intervening surface to reveal the black distant surface visible through the holes. even when the inducers are below the surface, which is Eleven subjects perceived neon colour spreading in a probably re¯ecting later neural processes. The reason for convex illusory triangle in the third stimulus. The pacmen earlier misinterpretation and limited neural modelling is were described as holes or funnels into which the apices the experimental restriction of each set of inducers to a penetrated, while remaining visible. Otherwise the convex single depth plane. surface was located in front of the white ®eld bearing the Additionally, our experiments showed no neon spreading contours of the holes. Two subjects (S.R. and S.S.) reported when inducers were perceived behind the intervening a curved triangle, partly occluded by a white plane so that surface appearing opaque. These results indicate that neural only apices were visible through the holes in the white computation of transparency/opaqueness modulates the opaque plane, and no neon colour spreading. construct of early processes. In the fourth stimulus twelve subjects reported a concave Black inducers generate a glowing white surface, that is, triangle, partly occluded by a white surface, with apices brightness enhancement in plain illusions. The glowing in visible since they reached through holes in the white neon ®gures could also be due to this brightness enhance- surface, and no neon colour spreading. Hence, the apices ment generated by luminance contrast between inducers and of the triangle appeared above the white surface whereas the surround. The use of the same basic neural mechanisms [1] rest of the triangle was occluded by it. One subject (S.S.) is also supported by the similar three-dimensional appear- reported that a concave triangle with neon colour spreading ance in curved Kanizsa ®gures with (present experiments) was visible through the white surface which was partly and without [2] colour spreading. transparent. Anatomical, neurophysiological and psychophysical When the tip of the upper sector and the horizontal edges evidence [5,8,10,12,18] suggests that in early vision illusory of the lower sectors were further away from the observer contours, illusory surfaces and neon spreading are processed than the rest of the con®guration (the ®rst and third stimu- in parallel. However, there is evidence that ®lling-in takes lus), a convex illusory triangle with neon colour spreading place in V1 [7], illusory contours are processed in V2 [15], was perceived by 12 (®rst stimulus) and 11 (third stimulus) and V4 is the main colour area in primates [19]. Serial 122 M. Liinasuo et al. / Neuroscience Letters 281 (2000) 119±122 processing is also a key element in the neural computation sentation in the visual system. Cogn. Brain Res., 5 (1996) suggested by Marr [9]. Viewing our stimuli monocularly 97±104. shows two-dimensional illusory contours and surfaces as [8] Livingstone, M.S. and Hubel, D.H., Psychophysical evidence for separate channels for the perception of form, well as neon spreading in each ®gure. However, stereo color, movement, and depth. J. Neurosci., 7(11) (1987) fusion radically alters this precept and even removes neon 3416±3468. colour and illusory con®guration. Our results agree with [9] Marr, D., Vision: a computational investigation into the serial or parallel processing of 2-D illusory con®gurations human representation and processing of visual informa- and neon colour spreading but suggest that depth processing tion, Freeman, New York, 1982, pp. 397. [10] Nakayama, K., Shimojo, S. and Ramachandran, V.S., Trans- and transparency follow these early neural processes. parency: relation to depth, subjective contours, luminance, and neon color spreading. Perception, 19(4) (1990) 497±513. Marja Liinasuo was supported by The Academy of [11] Soriano, M., Spillmann, L. and Bach, M., The abutting grat- ing illusion. Vis. Res., 36(1) (1996) 109±116. Finland and by The Alfred Kordelin Foundation, Ilpo [12] Van der Zwan, R. and Wenderoth, P., Psychophysical Kojo was supported by The Academy of Finland, and evidence for area V2 involvement in the reduction of Jukka HaÈkkinen by University of Helsinki, The Finnish subjective contour tilt after-effects by binocular rivalry. Cultural Foundation, The Alfred Kordelin Foundation, and Vis. Neurosci., 11(4) (1994) 823±830. Emil Aaltonen Foundation. [13] van Tuijl, H.F.J.M., A new visual illusion: neon-like color spreading and complementary color induction between subjective contours. Acta Psychol., 39 (1975) 441±445. [1] Bressan, P., Mingolla, E., Spillmann, L. and Watanabe, T., [14] Varin, D., Fenomeni di contrasto e diffusione cromatica Neon color spreading: a review. Perception, 26 (1997) 1353± nell'organizzazione spaziale del campo percettivo. Riv. 1366. Psicol., 65 (1971) 101±128. [2] Carman, G.J. and Welch, L., 3-Dimensional illusory [15] von der Heydt, R., Peterhans, E. and Baumgartner, G., Illu- contours and surfaces. Nature, 360(6404) (1992) 585±587. sory contours and cortical neuron responses. Science, 224 [3] Gregory, R.L. and Harris, J.P., Illusory contours and stereo (1984) 1260±1262. depth. Percept. Psychophys., 15 (1974) 411±416. [16] Ware, C., Coloured illusory triangles due to assimilation. [4] Grossberg, S., 3-D vision and ®gure-ground separation by Perception, 9 (1980) 103±107. visual cortex. Percept. Psychophys., 55(1) (1994) 48±120. [17] Watanabe, T. and Cavanagh, P., Transparent surfaces [5] Hubel, D.H. and Livingstone, M.S., Segregation of form, de®ned by implicit X junctions. Vis. Res., 33(16) (1993) color, and stereopsis in primate area 18. J. Neurosci., 2339±2346. 7(11) (1987) 3378±3415. [18] Watanabe, T. and Sato, T., Effects of luminance contrast on [6] Kanizsa, G., Quasi-perceptual edges in homogeneously color spreading and illusory contour in the neon color stimulated ®eds (Walter Gerbino, Trans.). In S. Petry and spreading effect. Percept. Psychophys., 45(4) (1989) 427± G.E. Meyer (Eds.), The Perception of Illusory Contours, 430. Springer-Verlag, New York, 1987, pp. 40±49. [19] Zeki, S. and Shipp, S., The functional logic of cortical [7] Komatsu, H., Murakami, I. and Kinoshita, M., Surface repre- connections. Nature, 335(6188) (1988) 311±317.