Illusory-Contour Figures Prime Matching of Real Shapes

Perception, 2009, volume 38, pages 1118 ^ 1131 doi:10.1068/p6118 Illusory-contour figures prime matching of real shapes Anna Barlasov-Ioffe, Shaul Hochstein Institute of Life Sciences and Neural Computation Center, Hebrew University, Jerusalem, 91904, Israel; e-mail: [email protected] Received 29 June 2008, in revised form 6 December 2008; published online 3 August 2009 Abstract. We investigated explicit and implicit properties of the internal representation of illusorycontour figures by studying potential priming effects of this representation. Using a primed matching paradigm (Beller 1971, Journal of Experimental Psychology 87 176 ^ 182), we found that illusory `Kanizsa' squares and triangles prime later matching of the same shapes, respectively, and not of the alternative shape. This priming effect is present despite the use of an illusory figure as a prime and real shapes as tests. To determine whether implicit processing mechanisms sufficiently induce a representation of the illusory shape so that it can lead to this priming effect, we used a novel method of presentation of the inducing pattern, based on Rock and Linnet's (1993, Perception 22 61 ^ 76) method for separating (implicit) retinal and (explicit) world-coordinate images. Presence of the implicit retinal image is confirmed by its producing an afterimage. While the retinal image is only implicitly produced by the inducing pattern of pacmen, it is nevertheless available for real-shape match priming. We conclude that Kanizsa-type inducer patterns are processed implicitly until formation of illusory-figure shapes. These are represented at relatively high cortical levels, and shape-matching priming must occur here, too. These results are consistent with the claim of the reverse hierarchy theory that bottom ^ up processing is generally implicit and that conscious perception originates at high cortical levels. 1 Introduction The phenomenon of illusory contours has been a subject of substantial interest and intensive research since its first presentation by Schumann more than a century ago (Schumann 1900). Illusory contours, also called subjective figures or illusory surfaces, are a case where we see lines, borders, surfaces, hues, and textures that are not physically present. Petry and Meyer (1987) define illusory contours such that produce a sense of (i) a bounded surface which has some property that differentiates it from a surrounding surface; (ii) a boundary (contour) around this surface; and (iii) the boundary and the surface that connect or continue through discontinuities of the inducing patterns. While this definition includes the notion of `a sense of ...', it is conceivable that some of the processes underlying this perception can be implicit or carried out without focused attention. If this were the case and these implicit processes are activated even when there is no explicit percept of the figure, then perhaps some figure properties may be available and affect performance of subsequent tasks. Thus, we now ask if it is possible that illusory figures are constructed by our visual system without our awareness. If so, to what degree and level of processing do these implicit mechanisms proceed? Implicit perception of illusory contours has been studied before only in spatial-neglect patients who were deprived of the explicit experience of seeing illusory-contour figures, as also real figures, but responded above chance to indirect measures of perception (Vuilleumier and Landis 1998; Vuilleumier et al 2001). To assess the possibility that at least some of the properties of the illusory figure (for example, its shape) can be processed implicitly in normal subjects, we previously tested illusory-figure detection and figure-shape discrimination simultaneously. We found a double dissociation between detection of the illusory contour figure and discrimination of its shape (Barlasov-Ioffe and Hochstein 2008). This result suggested that there Illusory-contour figures prime matching of real shapes 1119 may be separate processes leading to the explicit illusory-figure experience and to knowledge of its shape. We now ask if there can be a case where both the presence of the illusory figure and the representation of its shape are not perceived explicitly, but still there is an unconscious effect resulting from these processes. This unconscious effect cannot be measured directly (ie by asking questions regarding the illusory figure) but rather must be discerned by a related task that does not involve judgments concerning the induced shape or even its individual inducers. To this end, we adopted for our experiments a primed matching paradigm. This method, originally proposed by Beller (1971), depends on the finding that the response that two letter shapes are the same is speeded by previous exposure to the same letter shape. This paradigm was extended to non-letter shapes in studies of occluded-object completion (Sekuler and Palmer 1992) and perceptual grouping (Razpurker-Apfeld and Kimchi 2007). We used one of the most famous illusory contours, a Kanizsa figure, where a configuration of `pacmen' induces a clear illusory shape (Kanizsa 1979)öin this case either a triangle or a square. In experiment 1 we used this inducing pattern as a priming stimulus for a matching task. We tested for a priming effect when subjects report match or non-match for a pair of figures that are luminance-defined (ie real) triangles or squares (expecting a ``same'' response in either case), or one of each (expecting a ``different'' response). According to the rationale of the primed matching paradigm, when objects in the test pair are similar to the previously seen prime, correct ``same'' responses should be faster than when test objects are dissimilar from the prime. No priming effect is expected for ``different'' responses (Beller 1971). Since the illusory figure is defined by a surface with brightness contrast, we used brightness-defined figures also for the matching test. If figure integration that allows creation of the illusory triangle figure precedes the source of figure-match priming, then we should find priming for Kanizsa triangles as one does for real triangles (figure 1a). If, on the other hand, the site of figure integration prime expected response RT (a) (b) (c) (d) Figure 1. Application of the primed matching paradigm to the study of illusory-contour figure perception. The RT for ``same triangle'' response is expected to be shorter than that for ``same squares'' when primed by a Kanizsa triangle (a) and longer when primed by a Kanizsa square (b). If the priming effect is indeed based on the global priming illusory shape, then no speeding of the correct ``same'' response is expected when the priming stimulus is a pattern of rotated inducers (c) and (d). Rotated inducers alone could have a priming effect if priming were based on physical attributes of the inducing pattern (number, position, and/or angle of opening of the pacmen). 1120 A Barlasov-Ioffe, S Hochstein is higher than that of the priming mechanism, then priming should be absent in the Kanizsa case. It is an intriguing challenge to introduceöin normal subjectsöan illusory-contourinducing pattern in such a way that the viewers will not be aware of the induction, but their visual system will receive sufficient information for creating a figure. To solve this problem, in experiment 2 we varied the method of prime presentation. Here we propose a novel method of presenting a Kanizsa-type inducing pattern that allows creating at the same time separate images in retinal and world (screen) coordinates. This is an extension of the method proposed by Rock and Linnet (1993) for establishing whether perception of a shape is determined by its retinal image. The idea of their experiment was to break the image into components, and present these components separately and sequentially, while subjects track a moving fixation target between presentations. The absolute screen positions of the components determine what image is composed on the screen, but the relative position of the components with respect to the fixation target determines what image is composed on the retina. We were inspired to attempt this presentation method by our successful replication of the Rock and Linnet (1993) experiment and their result that subjects report perceiving a line-drawing image composed on the screen and not that on the retina (unpublished results). Thus, presenting the components of the inducing pattern (pacmen) sequentially with a moving fixation allows the composite inducing pattern to be created either in screen or in retinal coordinates. According to the previous results with line drawings, the inducing pattern composed in retinal coordinates is not expected to be perceived by the observer. We used a subsequent afterimage test to compensate for the eye movements and ensure the actual presence of the inducing pattern on the retina (Evans and Marsden 1966; Rock and Linnet 1993). 2 Experiment 1 Our first objective was to determine whether previous exposure to a Kanizsa-type illusory figure can prime the matching of real shapes. In terms of the primed matching paradigm, will ``same'' responses to the pair of triangles be faster than ``same'' responses to the pair of squares, if preceded by a Kanizsa-triangle prime; and vice versa, will ``same'' responses to the pair of squares be faster, if primed by a Kanizsa square (figures 1a and 1b)? If so, does this kind of priming depend on the induced shape, rather than only on the number of inducers and their orientations (figures 1c and 1d)? The selective and symmetric effects of illusory-triangle and illusory-square primes on the identical sets of test pairs will indicate a generalising aspect of priming, in that illusory figures prime the matching of their real correlates. 2.1 Methods 2.1.1 Experimental design. Subjects viewed a priming stimulus followed by a pair of test figures, as demonstrated in figure 2. Their task was to decide rapidly whether the two test figures were the same as each other or different. The test figures were either two triangles, or two squares, or a triangle and a square (with the triangle appearing on the right or on the left). Response reaction times (RTs) were measured from the moment the test figures appeared on the screen. Five types of prime conditions were used: no prime; two types of Kanizsa-illusoryfigure inducing patterns (triangle and square); and two types of rotated-inducer patterns (inducer number, position, and angle of openingöbut not orientationöcorresponding to those of the patterns inducing a triangle and a square, respectively). The `no-prime' case served as a baseline and control for any bias the subjects may have for triangles or squares. The rotated-inducers prime served to control for the possible effect of inducer number, position, and angle of opening. Illusory-contour figures prime matching of real shapes 200 ms 1121 test until response Pacman presentation Fixation 180 ms 100 ms Figure 2. Experiment 1: schematic representation of trial time course. After initial fixation (180 ms), a priming stimulus (eg a Kanizsa square) appeared for 200 ms. Following a 100 ms interval, a test pair of real shapes appeared until the subject responded, reporting if the two shapes were the same as each other or different (in this example, unlike the examples in figure 1, the correct answer is ``different''). For each trial, the prime was presented for 200 ms, followed, after a 100 ms interstimulus interval (ISI), by a test stimulus which remained on the screen until the subject's response (figure 2). The 5 prime type63 test type trials were presented once each, in pseudorandom order, for each of twenty blocks of trials (a mixed design). Overall, each subject performed one session of 300 trials. 2.1.2 Stimuli and setup. The size of the induced illusory figures was identical to the size of their luminance-defined (real) correlates in the test pairs. The sides of the triangle and square were 3.45 deg and 2.8 deg visual angle, respectively, while the radius of the pacmen was fixed at 0.7 deg. Thus, the support ratio (the ratio of the luminance-defined contour to the overall border) was 0.4 for the illusory triangle and 0.5 for the illusory square. The sizes of the triangle and the square were chosen to enable the centres of the pacmen to be always placed at the same distance from the central fixation (ie on a hypothetical circle 4 deg in diameter, circumscribing both of the shapes). While the prime stimuli were presented at the centre of the screen, the test shapes appeared on either side of the fixation, occupying together 8.8 deg63 deg. The pacmen were white on a grey background, inducing dark illusory figures. The test pair figures were darker than the background. Subjects were seated at a distance of 80 cm from the screen. Stimuli were presented in a dark room, on a 17 inch CTX computer monitor on which was affixed a circular aperture 16.7 deg in diameter, in order to disguise the square contour of the screen. The monitor was set at minimum brightness and maximum contrast to avoid an afterimage of the inducing pattern. 2.1.3 Subjects. Subjects were nine undergraduate and graduate students with normal or corrected-to-normal vision. Subjects were naive to the purpose of the experiment and were reimbursed for participation. 2.1.4 Data analysis. To assess the magnitude of the priming effect, taking into account our subjects' individual differences in baseline RTs, for each subject we calculated the normalised relative priming effect as: ‰RT squares† ÿ RT triangles†Š= 12 ‰RT squares† ‡ RT triangles†Š . This measure is expected to be positive for cases of primed matching of triangles, reflected in shorter RTs for ``same triangle'' responses, and negative for cases of primed (shorter RT) ``same square'' responses. We present this normalised priming effect averaged across subjects. 2.2 Results To evaluate the priming effect, we measured subjects' RTs in the matching task, comparing the correct ``same'' response RTs for the case of the two triangles test pair 1122 A Barlasov-Ioffe, S Hochstein (same-triangle response) and the case of the two squares test pair (same-square response), following a certain type of prime. The objective was to see whether the illusory triangle prime would induce faster same-triangle than same-square responses, and the illusory square prime would induce faster same-square than same-triangle responses. The error rate of the matching task was very low (3%  1% on average). Mean RTs for correct same-triangle and same-square responses in different prime type trials are presented in figure 3a with more detailed data (and statistical measures) shown in table 1. 0.15 no prime rotatedtriangle inducers rotatedsquare inducers Kanizsa triangle Kanizsa square RT=ms 525 500 475 Normalised priming effect 550 0.10 0.05 0.00 ÿ0.05 ÿ0.10 450 ÿ0.15 triangles squares Test shapes Kanizsa triangle (a) (b) rotated- no prime rotated- Kanizsa triangle square square inducers inducers Prime type Figure 3. Results of experiment 1 (N ˆ 9). (a) Mean correct ``same'' response RTs, following presentation of various prime types. (b) Normalised priming effect, calculated as the difference between correct same-square and same-triangle response RTs, divided by their average RT. Filled bars represent illusory-figure prime trials, empty bars represent non-inducing rotated `pacmen' pattern or no-prime trials. Above-zero effect indicates that triangles are matched faster than squares; below-zero effects mean squares are matched faster than trianglesöa priming effect in the opposite direction. ** Indicates result is significantly different than zero at p 5 0:01 level. Table 1. Correct ``same'' response RTs for experiment 1 (N ˆ 9). Significant differences between mean RTs for correct same-triangle and same-square responses, as found by paired t-test, are marked by asterisks. ** Signifies p 5 0:01. Standard errors are given in parentheses. Prime Illusory triangle Rotated-triangle inducers No prime Rotated-square inducers Illusory square t-test Response RTs/ms same-triangles same-squares 479 502 546 511 505 508 504 545 506 469 (21) (23) (24) (27) (25) (23) (24) (24) (27) (23) 0.001** 0.386 0.347 0.197 0.001** When no prime appeared before the test (figure 3a, X), the same-triangle and samesquare responses were of similar speeds (paired t-test, p ˆ 0:347, see table 1), establishing a baseline and indicating that our subjects did not have a predisposition to match one of the test figures faster than the other. The overall faster response with a prime may be due to the prime presentation eliciting expectancy for imminent test stimuli. In trials with an illusory triangle prime, same-triangle responses appear faster than same-square responses (figure 3a, black triangles; p ˆ 0:001), indicating a priming effect of the illusory triangle on matching real triangles. On the other hand, in trials with an illusory square as the prime, same-square responses were speeded relative to same-triangle responses (figure 3a, black squares; p ˆ 0:001). This, in turn, indicates that real-square matching is primed by an illusory square. The reciprocal priming effect is depicted by the crossed (X-like) pattern of performance in figure 3a. Illusory-contour figures prime matching of real shapes 1123 Considering the possibility of inducer number, position, or opening angle being responsible for the effect, we looked at trials where rotated inducers were used as primes. The data indicate that neither rotated triangle nor rotated square inducer patterns caused a significant speeding of the respective ``same'' responses (figure 3a, white empty triangles, p ˆ 0:386; white empty squares, p ˆ 0:197). This result supports the idea that it is the global induced illusory figure that primes the matching of the analogous real shapes. The mean normalised priming effect is presented in figure 3b. Positive values indicate that same-triangle responses were faster than same-square responses, while negative values indicate the opposite (see section 2.1). The main finding of experiment 1 is a selective and symmetric priming effect of illusory triangle and illusory square on the matching of corresponding real shapes. Together with the result that there was no such effect for the rotated-inducer patterns, this finding supports the conclusion that matching is primed by the shape of the induced illusory figures and not by the number, position, or angle of the openings of individual inducers. This suggests that this priming takes place at higher cortical levels where such illusory shapes are thought to be represented, above lower levels where only the inducer patterns are represented. We conclude that primed matching can be used as a tool for investigating higher-level mechanisms in the perception of Kanizsa-type illusory figures. 3 Experiment 2 The goal of this experiment was to study the priming effect of illusory contour figures, induced either in retinal or in screen coordinates, on matching of the corresponding real shapes. To this end, we varied the method of presentation of the priming stimulus. Besides the conventional method used in experiment 1 (referred to here as simultaneous presentation condition), we presented the inducing pattern of pacmen (as well as a non-inducing pattern of rotated pacmen) as a temporal sequence of individual pacmen, each in its appropriate location (condition called here sequential). To achieve separate retinal and screen images, we moved the fixation between the pacmen presentations (condition called here sequential with moving fixation). The first two presentation conditions were used for replication of the results of experiment 1 and to control for sequential pacmen presentation, respectively. 3.1 Methods 3.1.1 Experimental design. As in experiment 1, subjects viewed a priming stimulus followed by a pair of test figures that were either two triangles, two squares, or a triangle and a square. Their task was to decide rapidly whether the two test figures were the same as each other or different. Response reaction times (RTs) were recorded, from the moment the test figures appeared on the screen. We used three types of priming stimulus presentation: simultaneous, sequential with stationary fixation, and sequential with moving fixation (priming conditions are described in detail in figure 4), in a mixed design, with 480 trials divided into two sessions. The 10 prime-type63 test-type trials were pseudorandomly ordered for each of the sixteen blocks of 30 trials. In the simultaneous prime presentation condition (figure 4a), four types of prime were used: no prime (fixation only), real (luminance-defined) triangle, Kanizsa triangle, and a non-inducing pattern with rotated inducers. The real priming stimulus was similar in brightness to the test pair figures. The trial time course was identical to that in experiment 1. The prime was presented for 200 ms, followed, after a 100 ms ISI, by a test stimulus which remained on the screen until the subject's response. 1124 A Barlasov-Ioffe, S Hochstein Fixation position/trajectory Prime components temporal presentation simultaneous 200 ms Pacman presentation Fixation 180 ms (a) Composite prime image screen retina test until response 100 ms sequential 70 ms Pacman presentation Fixation 533 ms (b) 70 ms 533 ms 70 ms 533 ms test 100 ms 64 (c) (d) 70 ms Pacman presentation 100 ms Fixation stopped Fixation moving 533 ms 70 ms 100 ms 533 ms 70 ms 100 ms 533 ms test 100 ms 800 ms return to centre 64 Figure 4. Experiment 2. Prime presentation temporal sequences and resulting composite prime images. The prime was presented either simultaneously (a), ie all components at the same time, or sequentially [(b) ^ (d)], where each pacman appeared on the screen separately, one after the other, with a lengthy interval between them. When the fixation remained stationary [(a) and (b)], the resulting image on the screen matched that on the subject's retina. When subjects fixated a cross that moved between pacmen presentations, the pacmen were placed so that a triangle figure could be induced in either retinal (c) or screen (d) coordinates. In this case, each trial began with a fixation cross completing one circle without presentation of inducers. During the next four circles the inducing (or rotated) pacmen were presented, sequentially. In each case, the fixation cross was stopped for 100 ms to allow the eyes to stabilise, followed by a 70 ms presentation of the pacman. After completing the last circle, the fixation cross was returned to the middle of the screen (800 ms), where, after another 100 ms, the test figures were presented (on either side of the fixation cross). Four types of priming stimuli were used in condition (a): no prime, real triangle, Kanizsa triangle, and rotated pacmen pattern. In the other conditions there were two types of prime: Kanizsa triangle and rotated pacmen patterns. For demonstration purposes, the figure shows only examples with Kanizsa triangle primes. In the sequential prime presentation with stationary fixation condition, we used two types of prime: a Kanizsa triangle and a non-inducing pattern with rotated inducers. The prime was presented sequentially, ie each inducer was shown separately while the subject fixated a cross at the centre of the screen (figure 4b). Each inducer was presented for 70 ms with an interinducer interval of 530 ms, during which only the fixation cross was on the screen. The prime consisted of four repetitions of the three sequential inducer presentations. Illusory-contour figures prime matching of real shapes 1125 In the sequential prime presentation with moving fixation condition (figures 4c and 4d), subjects tracked a moving fixation cross as it moved in a circular trajectory. The inducing pattern of pacmen (or the non-inducing pattern of rotated inducers) was presented sequentially, as if at the corners of a triangleöeither in retinal or in screen coordinates, as follows. To create a retinal image of an inducing pattern (figure 4c), each pacman was shown (separately and sequentially) at the centre of the screen when the subject's eyes were at specific positions on the fixation trajectory. These positions were chosen so that the inducersöthough appearing always at the centre of the screen ^ would appear on the subject's retina at the corners of an imaginary composite triangle. The fixation cross circled in a counterclockwise direction; it stopped at 608, 1808, and 3008, relative to the top of the circle öeach time for 170 ms, first for 100 ms without any additional stimulation, followed by inducer presentation for 70 ms. Then the fixation cross motion was continued. Motion periods between stops were 533 ms (angular speed  2258 sÿ1). This method of presentation allowed an inducing pattern of pacmen (or non-inducing pattern of rotated inducers) to be constructed on the subject's retina, while creating a rotating pacman image in the centre of the screen. To create a screen image of an inducing pattern (figure 4d), each pacman appeared at the position of the fixation cross when it stopped at 1208, 2708, and 3608, relative to the top of the circle. In this way, an inducing pattern of pacmen (or non-inducing pattern of rotated inducers) was constructed on the screen, while creating overlaid pacman images on the retina. It should be noted that neither in the retinal nor in the screen condition did subjects report experiencing an illusion of a Kanizsa triangle. In the screen condition subjects reported perceiving a fragmented pattern of pacmen as well as being able to infer that had all the inducers appeared at the same time they would have expected to have seen the illusion. In the retinal condition they were unaware even of the inducing pattern itself (and perceived, instead, a rotating pacman in the centre of the screen). 3.1.2 Stimuli and setup. The experimental setup and the stimulus size were the same as in experiment 1. The composite images of priming stimuli in the sequential conditions were the same size as the corresponding priming stimuli in the simultaneous condition. The circular trajectory of the moving fixation was 4 deg in diameter. Subject's head position was stabilised by chin-and-forehead rest. 3.1.3 Subjects. Subjects were seven undergraduate and graduate students with normal or corrected-to-normal vision. Subjects were naive to the purpose of the experiment, and were reimbursed for participation. 3.2 Results We measured subject RT in the matching task, comparing the RT of same-triangle responses with that of same-square responses. The same subjects performed all the tasks in a mixed, within-subjects design. The error rate of the matching task was 2%  1% on average. Mean RTs for the correct ``same'' responses are presented in the graphs of figures 5a to 5d and more detailed data are shown in table 2. Figure 5a depicts a comparison of the mean RTs for correct same-triangle (left) and same-square (right) responses with different prime types in the simultaneous prime presentation case. Matching of triangles was significantly faster than that of squares when primed by a real (grey triangle symbols) or a Kanizsa triangle (black triangles; p 5 0:01; see table 2). There was no such priming effect in the case of rotated inducers (open triangle; p ˆ 0:260), indicating that it was that triangular shape and not merely the position of the inducers that affected the matching. There was also no difference 1126 A Barlasov-Ioffe, S Hochstein Simultaneous prime presentation stationary fixation Sequential prime presentation stationary fixation moving fixation retinal image screen image 550 500 no prime rotatedtriangle inducers 475 Kanizsa triangle RT=ms 525 real triangle 450 triangles squares triangles squares Normalised priming effect (a) (b) triangles squares triangles squares Test shapes (c) (d) 0.15 0.10 0.05 0.00 ÿ0.05 no prime real Kanizsa rotatedtriangle triangle inducers (e) Kanizsa rotatedtriangle inducers Prime type (f) Kanizsa rotated- Kanizsa rotatedtriangle inducers triangle inducers (g) (h) Figure 5. Experiment 2 results (N ˆ 7). Prime presentation conditions are indicated on top with the example images of the Kanizsa triangle. (a) ^ (d) Represent mean RTs for correct ``same'' responses following different prime types. Data symbols represent types of prime. Significant difference between same-triangle and same-square response RTs is marked by asterisks (*p 5 0:05, **p 5 0:01). (e) ^ (h) Represent normalised priming effect of different types of prime on the shape-matching test calculated as the difference between correct same-square and same-triangle response RTs, divided by their average RT (see section 2.1). When above zero, triangles are matched faster than squaresöindicating a priming effect (* p 5 0:05). A priming effect is present for all stationary-fixation triangle primes, whether real or illusory, simultaneous or sequential, but for moving-fixation primes only for that inducing a retinal-coordinate triangle. Table 2. Correct ``same'' response RTs for all conditions (N ˆ 7). Significant differences between mean RTs for correct same-triangles and same-squares responses, as found by paired t-test, are marked by asterisks (*p 5 0:05, **p 5 0:01). Standard errors are given in parentheses. Fixation Stationary Presentation simultaneous sequential Moving sequential (retina) sequential (screen) Prime t-test Response RTs/ms same-triangles same-squares no prime real Kanizsa rotated Kanizsa rotated 536 472 500 517 508 518 (19) (22) (18) (25) (28) (27) 537 514 521 521 530 516 (21) (27) (23) (24) (32) (28) 0.443 0.009** 0.008** 0.260 0.012* 0.438 Kanizsa rotated Kanizsa rotated 461 472 464 480 (20) (16) (16) (21) 477 471 461 470 (21) (27) (20) (20) 0.022* 0.495 0.307 0.171 Illusory-contour figures prime matching of real shapes 1127 between matching speed for squares and triangles without a prime (X symbols; p ˆ 0:443), establishing a uniform baseline for the effects. Though the priming effect was as significant with a Kanizsa-type prime as with a real-triangle prime, the magnitude of the priming effect was different (figure 5e). The real-triangle prime produced the largest effect, as would be expected as the result of this priming stimulus being identical to the triangles in the test pair. Note that the retinal position of the prime did not coincide with that of the test-pair shapes, but appeared between them. Thus, the priming effect on the match response is necessarily cognitive [see Imber et al (2005) for a similar retinal-position independent masking effect also seen as cognitive]. In the sequential condition with stationary fixation, the pacmen composing the inducing (or non-inducing) prime pattern were separated temporally. Thus, short-term (working) memory would be involved in this case not only for retaining the prime stimulus during the subsequent matching task (and perhaps for performing the matching task itself), but also for perceiving the prime stimulus in its entirety. Figure 5b presents RTs measured for same-triangle (left) and same-square (right) responses. The matching was affected by the Kanizsa triangle prime (black triangle symbols) and not by rotated inducers (open triangles): the reaction time for the same-triangle response was significantly shorter than that for the same-square response (table 2, p 5 0:05). In the sequential prime presentation condition with moving fixation, the pattern of priming pacmen was composed either in retinal or screen coordinates (with overlapping, rotating pacmen in the other coordinates; see figures 4c and 4d). The RT for the same-triangles response was significantly shorter than for the same-squares response only when primed by a Kanizsa inducing pattern composed on the retina (figure 5c, black triangle symbols; p 5 0:05). The inducing as well as the non-inducing pacmen pattern in screen coordinates did not produce a priming effect on the matching test (figure 5d). Note that subjects were not aware of the retinal coordinate pattern (inducing triangle or rotated pacmen), suggesting that the illusory triangle was induced even in the absence of explicit experience of the inducing pattern, and that this information was available for later priming of real-shape perception. In general, response mean RTs with moving fixation were shorter than with stationary fixation. This may be due to a difference in the experimental paradigm used for stationary and moving fixations. In the moving-fixation case, after completing the last circle, the fixation cross was returned to the centre prior to the test presentation. This could have been used as a cue for the upcoming test pair and could therefore have made the responses quicker. Surprisingly, the normalised relative priming effects for Kanizsa triangle primes in simultaneous, sequential stationary, and sequential moving retinal (black bars in figures 5e, 5f, or 5g, respectively) conditions are of similar magnitude. The results of experiment 2 fully replicate those of experiment 1, establishing a primed matching paradigm as a reliable method of investigating illusory contour figure perception. Additionally, the results support the notion of the original paradigm, stating that the effect depends on the similarity between the prime and the matched shapes. The real triangle prime produced a bigger effect than the illusory triangle. This could also be due to the separate representations of real and illusory-contour figures at higher cortical levels (Imber et al 2005). 4 Afterimage illusion To control for eye movements in experiment 2, after completing the experiment subjects performed an afterimage test. They viewed the sequentially presented Kanizsa triangle and square inducers with stationary and with moving fixation conditions. The sequence of these trials was identical to those in figures 4b and 4c, except that the 1128 A Barlasov-Ioffe, S Hochstein inducing cycle was repeated ten times, and pacmen were white on a black background. Overall, there were 2 shapes62 conditions (stationary and moving)62 repetitions in pseudorandomised order. As in experiment 2, the subject's head position was stabilised by a chin-and-forehead rest. After viewing each stimulus, subjects were instructed to look at a blank grey screen with a central fixation cross, and to report whether they obtained an afterimage. We were particularly interested in whether they would report an afterimage only of the inducers, or whether they would also perceive illusory figures. An afterimage illusory Kanizsa figure was reported in both stationary and moving fixation conditions. To quantify the clarity of the illusion in the afterimage, we asked subjects to scale the clarity of the afterimage illusory shape (square or triangle) from 1 to 5. Subjects were asked to base their judgment on comparison of the illusion perceived in the afterimage with the illusion perceived in free viewing of the inducing pattern on paper. There are clear afterimage and illusory triangle/square effects in the stationary fixation condition (figure 6, left). In the moving fixation condition (figure 6, right), the effect is weaker but still significant. We conclude that subjects were indeed able to reliably maintain fixation of the moving target in experiment 2, and that an inducing pattern was composed on their retina in this condition. time Illusion clarity rating 5 4 3 2 1 0 triangle square triangle square stationary fixation moving fixation Illusory-contour adapting stimulus Figure 6. Subjects' mean (N ˆ 7) ratings of the clarity of the illusory figure perceived while viewing the afterimage, comparing it to the clarity of the illusion perceived in free-viewing of inducing pattern on paper (5: full illusion; 1: no illusion). The adapting stimuli of two sequential presentation conditions were used: with stationary fixation and with moving fixation creating a retinal image. A schematic diagram of component presentation for the two conditions (stationary and moving) for both shapes (triangles and squares) is shown on top. It should be noted, once again, that in this moving fixation condition, there is only implicit information present about the inducing pattern of pacmen. The screen coordinate presentation itself does not actually have a figure in it, only a central rotating pacman. Nevertheless, when this information is perceived öas an afterimageöit produces the percept of a Kanizsa illusory triangle. Most of the subjects expressed their surprise seeing an inducing pattern and an illusory triangle as an afterimage, after not being aware of either of them while viewing the adapting stimulus. Thus, the recycling of implicit information, by means of an afterimage, creates a new conscious perceptöin this case, an illusion of a Kanizsa triangle or square. Illusory-contour figures prime matching of real shapes 1129 5 General discussion In this study we investigated properties of the internal representation of illusory contour figures. For this purpose, we measured the effect of illusory contour figures (among other control stimuli) on subsequent matching of corresponding real shapes. We found that modally completed illusory-contour figures, such as Kanizsa triangles and squares, have a significant priming effect on the matching of real contour triangle and square shapes, respectively. Rotated inducers do not prime real figures: thus, it is neither the positions of the inducers, nor the missing sector from each inducer that causes the priming, but rather the induced illusory figure. These findings suggest that this priming takes place at higher cortical levels where such illusory shapes are thought to be represented, above lower levels where only the inducer patterns are represented. As we mentioned above, priming in the primed matching paradigm depends on similarity between the prime and the matched shapes. If it is assumed that a stronger illusion is more similar to the luminance-defined shapes (in the matching test), then a larger priming effect is expected when using a larger support ratioöwhich enhances illusory contour strength (Shipley and Kellman 1992). However, in our experiment 1, while the support ratios for the illusory triangle and square are different, the priming effects are the same. It is possible that this kind of priming does not depend on the strength of the illusory contours, but on the presence of an illusory surface that are represented differentially in the brain (Stanley and Rubin 2003, 2005). We also found that when Kanizsa shapes are induced by sequential presentation of their inducers, they prime subsequent matching of real figures despite the fact that the inducing cycle of the sequential presentation of the pacmen was too long for perception (Kojo et al 1993) and the intervals between the inducers were too long for retention of the information in V1 (Takemoto and Ejima 1997). This inescapably leads us to the conclusion that cortical areas higher than V1 must be responsible for integration of the inducing pattern into a composite shape. This conclusion is in tune with fMRI and MEG findings of strong activation of higher-tier retinotopic areas by illusory contours (Mendola et al 1999; Halgren et al 2003). Nevertheless, consistent retinotopic input (ie pacmen positioned so that they would form an inducing pattern in retinotopic coordinates) is necessary for the illusory figures to prime the matching of real shapes, as demonstrated by the fact that inducing patterns in retinal and not in world coordinates produced the priming effect. One may seek an alternativeölower-level effectöinterpretation of our results based on well-documented activation of lower cortical areas such as V1 and V2 in response to illusory contours (von der Heydt et al 1984; Grosof et al 1993; Hirsch et al 1995; Seghier et al 2000; Ramsden et al 2001). However, the priming effect that we find appears to depend on illusory surface of particular shape rather than on the presence of illusory contours. For example, in experiment 1 same-triangle response RT in the matching task was unaffected by both Kanizsa square and rotated-triangle inducers as primes (figure 4a; table 1). Symmetrically, Kanizsa triangle and rotated-square inducers both did not shorten RT for same-square response. In other words, illusory contours of one Kanizsa shape did not affect the matching of another shape in the same way as rotated inducers that did not produce any illusory contours at all. Thus, mere illusory contours are insufficient for any priming effect on shape matchingöit depends on presence of an illusory surface of congruent shape, which is believed to be constructed in higher cortical areas. 5.1 Implicit illusory figure formation As found by Rock and Linnet (1993) and confirmed by our own replication of their experiment (unpublished), the composite image perceived by the viewer is the one that is sequentially composed on screen and not the one composed on the retina. 1130 A Barlasov-Ioffe, S Hochstein Although in neither moving fixation condition did our subjects report experiencing the illusion, when the inducing pattern was composed on the screen (figure 4d) they reported being able to infer that had all the inducers appeared at the same time they would have expected to have seen the illusion. Such inference was apparently insufficient for this kind of priming. One may expect such information to be sufficient for more cognitive tasks, such as naming or association. At the same time, when subjects were unaware of even the inducing pattern itself, as was the case with the inducing pattern being composed on the retina (figure 4c), matching of the real shapes was primed to the same degree as with simultaneous presentation of the inducing pattern. The fact that conscious perception of the illusion was not necessary for the priming effect suggests that the process of illusory-figure formation may be implicit until the level of representation of the induced shape. The phenomenon of subliminal visual priming has been described by Bar and Biederman (1998), who found that naming accuracy of previously unidentified objects increased after a second presentation. Their further studies suggested that subliminal visual priming is mediated by the human homologue of macaque V4 (Bar and Biederman 1999). TMS studies report parallel results concerning area V5/MT and priming of motion direction (Campana et al 2002). Moreover, the latter investigators found that perceptual priming was not affected by magnetic stimulation over striate or parietal cortex, suggesting that no back-projection to primary visual areas is necessary for priming. Hochstein and Ahissar (2002; see also Ahissar and Hochstein 1997, 2004) have suggested that the normal course of events in the visual system is implicit processing along the well-known hierarchy of cortical areas. Conscious perception is understood to begin only when information has passed through these implicit processing stages and reached the highest cortical areas, where concepts and categories are represented. Conscious perception of details within a scene is a reverse-hierarchy return to the lower areas when this information is already represented. This reverse-hierarchy return is carried out under top ^ down control. The present results are consistent with the reverse hierarchy theory model in that we have now shown that processing of illusory figures starts in a bottom ^ up implicit manner. Furthermore, the first conscious percept is of the image in world-centred coordinatesönot retinal coordinates. Thus, when the retinal image is different from the image in screen coordinates (as in our sequential condition with moving fixation) the attention-driven feedback to lower areas is ineffective for conscious perception of illusory contours. Finally, without such conscious perception, there is still implicit priming of future stimulus presentations (in the matching paradigm) allowing faster responses for figures that are conceptually the same as those appearing in the only place they can appearö on the retina. Acknowledgments. This study was supported by grants from the Israel Science Foundation of the Israel Academy of Sciences and Humanities, and from the US ^ Israel Bi-national Science Foundation (BSF). We thank Anne Treisman, Bob Shapley, Nancy Kanwisher, Dov Sagi, and Asher Cohen for enlightening discussions of this work. References Ahissar M, Hochstein S, 1997 ``Task difficulty and the specificity of perceptual learning'' Nature 387 401 ^ 406 Ahissar M, Hochstein S, 2004 ``The reverse hierarchy theory of visual perceptual learning'' Trends in Cognitive Sciences 8 457 ^ 464 Bar M, Biederman I, 1998 ``Subliminal visual priming'' Psychological Science 9 464 ^ 469 Bar M, Biederman I, 1999 ``Localizing the cortical region mediating visual awareness of object identity'' Proceedings of the National Academy of Sciences of the USA 96 1790 ^ 1793 Illusory-contour figures prime matching of real shapes 1131 Barlasov-Ioffe A, Hochstein S, 2008 ``Perceiving illusory contours: Figure detection and shape discrimination'' Journal of Vision 8(11):14, 1 ^ 15 Beller H K, 1971 ``Priming: effects of advance information on matching'' Journal of Experimental Psychology 87 176 ^ 182 Campana G, Cowey A, Walsh V, 2002 ``Priming of motion direction and area V5/MT: a test of perceptual memory'' Cerebral Cortex 12 663 ^ 669 Evans C R, Marsden R P, 1966 ``A study of the effect of perfect retinal stabilization of some wellknown visual illusions, using the after-image as a method of compensating for eye movements'' British Journal of Physiological Optics 23 242 ^ 248 Grosof D H, Shapley R M, Hawken M J, 1993 ``Macaque V1 neurons can signal `illusory' contours'' Nature 365 550 ^ 552 Halgren E, Mendola J, Chong C D R, Dale A M, 2003 ``Cortical activation to illusory shapes as measured with magnetoencephalography'' NeuroImage 18 1001 ^ 1009 Heydt R von der, Peterhans E, Baumgartner G, 1984 ``Illusory contours and cortical neuron responses'' Science 224 1260 ^ 1262 Hirsch J, DeLaPaz R, Relkins N R, Victor J, Kim K, Li T, Borden P, Rubin N, Shapley R, 1995 ``Illusory contours activate specific regions in human visual cortex: evidence from functional magnetic reasonance imaging'' Proceedings of the National Academy of Sciences of the USA 92 6469 ^ 6473 Hochstein S, Ahissar M, 2002 ``View from the top: hierarchies and reverse hierarchies in the visual system'' Neuron 36 791 ^ 804 Imber M L, Shapley R M, Rubin N, 2005 ``Differences in real and illusory shape perception revealed by backward masking'' Vision Research 45 91 ^ 102 Kanizsa G, 1979 Organization in Vision: Essays on Gestalt Perception (New York: Praeger) Kojo I, Linasuo M, Rovamo J, 1993 ``Spatial and temporal properties of illusory contours'' Vision Research 33 897 ^ 901 Mendola J D, Dale A M, Fischi B, Liu A K, Tootell R B H, 1999 ``The representation of real and illusory contours in human cortical visual areas revealed by fMRI'' Journal of Neuroscience 19 8560 ^ 8572 Petry S, Meyer G E, 1987 The Perception of Illusory Contours (New York: Springer) Ramsden B M, Hung C P, Roe A W, 2001 ``Real and illusory contour processing in area V1 of the primate: a cortical balancing act'' Cerebral Cortex 11 648 ^ 665 Razpurker-Apfeld I, Kimchi R, 2007 ``The time course of perceptual grouping: The role of segregation and shape formation'' Perception & Psychophysics 69 732 ^ 743 Rock I, Linnet C M, 1993 ``Is a perceived shape based on its retinal image?'' Perception 22 61 ^ 76 Schumann F, 1900 ``Beitra«ge zur Analyse der Gesichtswahrnehmungen. Erste Abhandlung. Einige Beobachtungen u«ber die Zusammenfassung von Gesichtseindru«cken zu Einheiten'' Zeitschrift fu«r Psychologie und Physiologie der Sinnesorgane 23 1 ^ 32 Seghier M, Dojat M, Delon-Martin C, Rubin C, Warnking J, Sedebarth C, Bullier J, 2000 ``Moving illusory contours activate primary visual cortex: an fMRI study'' Cerebral Cortex 10 663 ^ 670 Sekuler A B, Palmer S E, 1992 ``Perception of partly occluded objects: A microgenetic analysis'' Journal of Experimental Psychology: General 121 95 ^ 111 Shipley T F, Kellman P J, 1992 ``Strength of visual interpolation depends on the ratio of physically specified to total edge length'' Perception & Psychophysics 52 97 ^ 106 Stanley D A, Rubin N, 2003 ``fMRI activation in response to illusory contours and salient regions in the human lateral occipital complex'' Neuron 37 323 ^ 331 Stanley D A, Rubin N, 2005 ``Rapid detection of salient regions: Evidence from apparent motion'' Journal of Vision 5(9):4 Takemoto A, Ejima Y, 1997 ``Retention of local information in generation of subjective contours'' Vision Research 37 1429 ^ 1439 Vuilleumier P, Landis T, 1998 ``Illusory contours and spatial neglect'' NeuroReport 9 2481 ^ 2484 Vuilleumier P, Valenza N, Landis T, 2001 ``Explicit and implicit perception of illusory contours in unilateral spatial neglect: behavioural and anatomical correlates of preattentive grouping mechanisms'' Neuropsychologia 39 597 ^ 610 ß 2009 a Pion publication ISSN 0301-0066 (print) ISSN 1468-4233 (electronic) www.perceptionweb.com Conditions of use. This article may be downloaded from the Perception website for personal research by members of subscribing organisations. Authors are entitled to distribute their own article (in printed form or by e-mail) to up to 50 people. This PDF may not be placed on any website (or other online distribution system) without permission of the publisher.