Conscious perception preceding attention: the effect of Kanizsa illusion on the capacity of iconic memory and fragile VSTM

Conscious perception preceding attention: the effect of Kanizsa

illusion on the capacity of iconic memory and fragile VSTM

Klaudia Ambroziak

Abstract

Previous studies (Sligte, Scholte & Lamme, 2008) discern three stages of memory in visual information processing: 1. iconic memory with unlimited capacity, 2. a four seconds lasting fragile visual short term memory (VSTM) with a capacity that is at least a factor of two higher than the third stage which is 3. the robust and capacity-limited form of VSTM. It has already been shown that iconic memory and fragile VSTM contain a very rich representation of many more objects than are present in working memory. However, it remains an open question, whether iconic memory and fragile VSTM represent conscious forms of visual processing. The aim of this research was to investigate whether the representations stored in iconic memory and fragile VSTM have perceptual qualities that characterize conscious perception i.e. perceptual inference. To manipulate perceptual inference, we used a change detection task with the Kanizsa subjective contours illusion, where subjects perceive illusory contours of figures guided by inducers. As predicted, we found that in trials with Kanizsa subjective contours the capacity of iconic memory and fragile VSTM was significantly higher than in control trials in which the same physical features were present, but no illusory Kanizsa figures could be formed. The effect of the illusion was also larger for iconic memory and fragile VSTM than for working memory. These results show that iconic memory and fragile VSTM rely on perceptual qualities that characterize conscious vision and thus, in addition to working memory, might reflect conscious processing.

1. Introduction

How rich and detailed is our conscious vision? Even when confronted with a complex visual scene for a very brief moment, we often experience a richness of content. But how much of this complex scene can we consciously perceive? Is focusing attention necessary to have conscious representations with all its phenomenal aspects? Many scientists tend to associate consciousness with attention; some conflate these processes entirely (Posner, 1994, Merikle & Joordans, 1997), whereas others argue that without attention conscious perception cannot occur (Dehaene, Changeux, Naccache, Sackur & Sergent, 2006; de Gardelle, Sackur & Kouider, 2009; Kouider, de Gardelle, Sackur & Dupoux, 2010). Furthermore, it is often suggested that subjective report should be the primary criterion that can establish whether a percept is conscious or not (Dehaene et al. 2006; Weiskrantz, 1997).

Other researchers, however, propose that consciousness is not the same as attention, and that subjective reporting should not stand as a main criterion in deciding whether a percept is conscious or not. Block (1990, 2007) suggested a state of ‘phenomenal consciousness’ prior to global access: when we look at a complex visual scene, we experience a richness of content that goes beyond what we can report. Therefore, Block proposed a distinction between “phenomenal consciousness” and “access consciousness”. Similarly, Koch and Tsuchiya (2007) argue that attention and consciousness are distinct phenomena that need not occur together and that can be manipulated using distinct paradigms.

In a model proposed by Lamme (2003, 2006, 2010), attention is determined by the depth of processing, whereas consciousness relies on recurrent processing. Hence, two orthogonal divisions are proposed: un-attended perception (shallow processing) vs. attended perception (deep processing), and conscious perception (recurrent processing) vs. unconscious perception (feedforward processing). When combining the depth of the processing with the feedforward vs. recurrent processing, this model distinguishes four stages of processing. Stage 1 and 2 are characterized by a feedforward sweep and are consequently unconscious. Stage 3 consists of recurrent but unattended, and thus shallow, processing. In this stage the recurrent processing stays limited to lower brain areas and creates phenomenal consciousness. In stage 4 however, attention boosts the processing to be spread to the frontal areas, which deepens the recurrent processing and results in access consciousness.

How can we determine whether shallow processing of recurrent nature (stage 3) has the same perceptual qualities as deep recurrent processing (stage 4)? The implications of this theory are as follows: since stage 3 and stage 4 are characterized by recurrent processing, they both result in conscious percepts. However, in stage 3 recurrent processing only occurs in lower areas, which results in a non-reportable form of conscious vision. Therefore, it is not possible to directly assess its features.

Nevertheless, evidence has been found that the representation that is formed in stage 3 processing will remain in short term memory. Previous studies (Sligte, Scholte & Lamme, 2008) discern three stages of memory in visual information processing: 1) iconic memory with unlimited capacity, 2) a four seconds lasting fragile visual short term memory (VSTM) with a capacity that is at least a factor of two higher than the third stage which is 3) the robust and capacity-limited form of VSTM. Iconic memory has been thoroughly investigated for over 50 years and strong evidence supporting the notion of its very rich representation comes from studies using Sperling’s partial report paradigm (Sperling, 1960). In this classic task, subjects are presented with brief (50 ms) visual stimuli containing 3 arrays of letters. After disappearance of the stimulus subjects are given an auditory cue which indicates a specific line of letters from the initial display. Then subjects are asked to recall the letters from the cued line. On average subjects are able to report 3.3 out of 4 letters, when the cue is presented less than 1 second after the offset of the stimulus. Recall based on a cue which immediately follows the offset of the stimulus demonstrates the capacity of iconic memory. Performance in the partial report task can be regarded as a random sample of subject’s memory for the entire display: because subjects do not know which row will be cued for recall, and a cue appears only after a stimulus display has disappeared, it is inferred that iconic memory contains a representation of at least 9 letters of the whole stimulus display (12 letters). When there is no cue, subject can rely solely on working memory and report drops to 3-5 letters out of entire display (12 letters). These findings shows that capacity of iconic memory is much higher than capacity of working memory (representing stage 4 processing and thus access consciousness).

Previous research (Landman, Spekreijse & Lamme, 2003; Sligte et al. 2008; Sligte, Vandenbroucke, Scholte, & Lamme, 2010) suggests that not only iconic memory but also fragile VSTM contains a very rich representation of many more objects than are present in working memory. It was also shown that a decrease in attention greatly reduced the capacity of visual working memory, but had only a small effect on the capacity of fragile VSTM (Vandenbroucke, Sligte & Lamme, 2011). This provides evidence that different types of memory represent different stages of visual processing, and that representation in fragile VSTM can be formed independent of attention. However, it remains an open question, whether iconic memory and fragile VSTM are conscious types of memory.

Objects stored in fragile VSTM seem to be represented in a state where features are bound, figure-ground segregation has occurred, and where details of objects are discriminated (Landman, et al., 2003; Sligte et al. 2008; Sligte et al., 2010). Thus, fragile VSTM not only resembles iconic memory, but also has a lot of characteristics in common with working memory. It is therefore proposed that fragile VSTM incorporates information created in stage 3 processing and might represent phenomenal consciousness. How do we know if the representation in fragile VSTM has phenomenal aspects? Can we regard this stage of memory as conscious? The aim of the present research was to investigate whether the representations of unattended objects stored in fragile VSTM and sensory memory have perceptual qualities that characterize conscious perception i.e. perceptual inference. Conscious percepts may differ from the physical qualities of the stimulus and rely on meanings inferred from the context.

To manipulate perceptual inference we used the Kanizsa subjective contours illusion (Fig. 1), where subjects perceive illusory contours of figures guided by inducers. Previous studies (Harris, Schwarzkopf, Song, Bahrami & Rees, 2011) showed that when the inducers were masked from awareness, subjects did not experience illusory contours: their behavioral performance dropped to chance level compared to control trials with unmasked stimuli. Thus, perception of illusory contours depends on the awareness of the inducing elements. This suggests that conscious processing of spatial context is necessary for Kanizsa illusion to occur and high-level inferential mechanisms are involved. In this case, we assumed the following: If a brief presentation of a visual scene is

3

enough for perceptual inference to occur, the Kanizsa illusion should be perceived by subjects that were presented with the scene for only a brief moment (in a way that prevented from focusing attention on all objects). We asserted this criterion as an indication whether subjects were conscious or not.

To assess whether subjects were affected by the Kanizsa illusion we used a cued change detection task (see Materials and Methods). We then compared the capacity of iconic memory and fragile VSTM, with the capacity of visual working memory, which stores representations of attended objects formed in deep, recurrent processing (stage 4 in Lamme’s model). Consequently, the main question addressed in the first study was whether the Kanizsa illusion positively influenced subjects’ performance in iconic memory and fragile VSTM compared to working memory. In the second experiment we further explored the dependency of iconic memory and fragile VSTM on perceptual inference. We investigated how the strength of the illusion influenced subjects’ performance. To do this, we changed the size of inducing elements (see Fig. 4), and thus the support ratio (the ratio of the length of the real inducing contours relative to the total length of the illusory figure) of the illusory figures.

2. Experiment 1

In experiment 1 we aimed to compare the capacity of iconic memory, fragile VSTM, and visual working memory for figures with Kanizsa subjective contours (illusions) and figures with the same physical properties, but without the subjective contours being formed (controls). Subjects performed a change detection task including control and illusion memory arrays. We expected that it would be easier for subjects to perceive a change in the visual scene in trials with Kanizsa subjective contours compared to control trials in which no illusory Kanizsa figures could be formed. The assumption that working memory, which represents conscious processing, would benefit from the Kanizsa illusion, was not controversial. The prediction of this research, however, was that the illusion would also have an impact on fragile VSTM and sensory memory. This would suggest that iconic memory and fragile VSTM already contain perceptual qualities such as perceptual inference. We expected that the impact of illusion on fragile VSTM would be equal to or even bigger than the effect of the illusion on working memory, taking into account that the capacity of iconic memory and fragile VSTM is larger than capacity of working memory.

2.1 Materials and Methods

2.1.1 Subjects

Twenty-one students (15 females and 6 males, aged 19-51, average 25) with normal or corrected-to-normal vision participated in this experiment. Twenty of them (14 females and 6 males) passed the training (basic version of the change detection task, see Procedure) and participated in the experimental task. For their participation, subjects received course credits or monetary reward. All subjects gave their written informed consent to participate in this study. The experiment was approved by the local ethics committee of the department of Psychology of the University of Amsterdam.

2.1.2 Stimuli

In the basic version of the change detection task there was one set of stimuli which consisted of white rectangles orientated horizontally, vertically, 45° to the vertical, and 135° to the vertical. 4 Subjects were shown memory and test displays containing eight rectangles (1.6°x0.45°) placed radially within 2.29° from a red fixation dot. Cues consisted of a 3-pixel thick line (2° long). Stimuli were presented on a black background. In the actual experiment there were two possible sets of stimuli (see Fig. 1), which constituted an experimental and a control condition. Both sets included two figures (2°x3.15° each) and each of them consists of four, white pacmen.

In the first set (experimental condition, see Figure 1A) pacmen formed a figure with a Kanizsa triangle pointing outwards or inwards relative to the fixation dot. Support ratio (the ratio of the physically specified contours to the total contour length) of the Kanizsa triangle was 0,67. In the second set (control condition, see Figure 1B) pacmen formed non-Kanizsa figures with middle pacmen rotated inwards or outwards compared to the center of the figure. Subjects were shown memory and test displays containing eight figures from one of two possible sets, placed around the red fixation dot (at 2.29° distance). Two figures at the bottom and two figures at the top were orientated vertically, while two figures on the left side and two figures on the right side were orientated horizontally (see Fig.2). All stimuli were presented on a black background.

A: Kanizsa figures

B: Controls

Figure 1: Two sets of stimuli, which constituted an experimental (A) and a control (B) condition.

2.1.3 Task

The red fixation dot in the middle of the screen turned green for 1,000-ms to indicate the start of the trial. Then, a memory display containing eight figures appeared for 500-ms. Subjects were instructed to remember as many objects of this memory display as possible. During each trial, one figure was cued by a line pointing towards its location from the center of the screen to indicate which item was the one to report. After a blank interval in which no stimulation was provided, a test display was shown.

Subjects were asked to indicate by button press whether the cued figure was the same (50% of the trials) or different (50% of the trials) compared to the figure which was shown at the same location in the memory display. All of the other 7 figures were always the same between memory and test display. Test displays were present for 4,000-ms or until the subject made a response. On each trial subjects received auditory feedback about the correctness of their response.

5

Cues were introduced at different latencies during the trial; either 30 ms after off-set of the memory display (iconic cue), 1,000 ms after off-set of the memory display (retro-cue), or 100 ms after the on-set of the test display (post-change cue). The interval between memory and test display was 1,030 ms for the iconic condition, 2,000 for the retro-cue conditions, and 900 ms for the post-change cue conditions, while the interval between the cue and test display in iconic and retro-cue condition was 1,000 ms.

2.1.4 Procedure

First, we tested subjects on visual acuity and depth perception. Thereafter, to avoid confounds of learning rates which may differ between sensory and working memory, subject were trained (for a minimum of 4 and maximum of 12 blocks of 60 trials) on a basic version of the change detection task containing white rectangles instead of pacmen figures. After subjects had reached a performance level of 75%, they were trained on the actual experiment containing the Kanizsa figures for one block divided into four parts of 72 trials (12 trials per condition: cue-timing (3) × figures (2) x change/no change (2)) resulting in a total of 288 trials (25 minutes containing 3 short breaks). In the second session subjects first performed 12 trials (one trial per condition:

cue-Figure 2. Task design with Kanizsa figures presenting three types of trials: A) measuring

iconic memory with an iconic cue 30 ms after off-set of the memory display, (B) measuring fragile visual short-term memory with a retro cue 1,000 ms after off-set of the memory display, and (C) working memory with a post-change cue 100 ms after the on-set of the test display.

Memory display (500 ms)

A

B

C

timing (3) × figure (2) x change/no-change (2)) to get rid of start-up effects. Then, in the experimental task subjects performed 3 blocks divided into four parts of 72 trials (12 trials per condition: cue-timing (3) × figures (2) x change/no change (2)). Thus, the task consist of 144 trials in each condition (cue-timing (3) × illusory/control (2)), resulting in a total of 864 trials.

2.2 Results from Experiment 1

In the present study we aimed to compare the capacity of iconic memory, fragile VSTM, and visual working memory for two types of figures: the one containing Kanizsa subjective contours and the controls. We predicted that it will be easier for subjects to perceive a change in the visual scene in trials with Kanizsa subjective contours compared to control trials in which the same physical features were present, but no illusory Kanizsa figures could be formed. We expected that the effect of the illusion on iconic memory and fragile VSTM would be equal to or even bigger than the effect of illusion on working memory.

The results were analyzed using repeated measures ANOVA. As predicted by the theory, for all types of memory subjects’ performance was significantly higher for Kanizsa figures than for the controls (F(1,18) = 109.208 , p <.001) (see Figure 3). For both types of figures, performance decreased over memory conditions (F(2,17) = 55.89, p < .001) ), replicating findings that iconic memory has a larger capacity than fragile VSTM, and the capacity of fragile VSTM is larger than capacity of working memory. Furthermore, according to our expectation the effect of the illusion on iconic memory and fragile VSTM was even bigger than the effect of the illusion on working memory. Differences in capacity between Kanizsa figures and control figures for iconic memory, fragile VSTM and visual working memory were 14.8%, 13.6% and 9.4% respectively (F(2,17)= 4.951, p = .02). Since iconic memory and fragile VSTM had a benefit from Kanizsa illusion, this suggests that iconic memory and fragile VSTM already contain perceptual qualities such as perceptual inference.

Figure 3. Experiment 1 Results: Percentage correct on the Change Detection task for iconic memory,

fragile memory and working memory on the Kanizsa and Control figures. 0.5 0.55 0.6 0.65 0.7 0.75 0.8

Iconic Fragile Working

%C orre ct Memory type

Experiment 1 Results

7 2.3 Discussion of Experiment 1

In this experiment we showed that the Kanizsa illusion had a positive effect on subjects’ performance in a change detection task compared to control trials in which the same physical features were present, but no illusory Kanizsa figures could be formed. The illusion not only influenced working memory, which was a noncontroversial assumption, but also had an impact on the iconic memory and fragile VSTM capacity. The effect of the illusion was even bigger for iconic memory and fragile VSTM than for working memory. This suggests that iconic memory and fragile VSTM already contain perceptual qualities such as perceptual inference. These results show that the brief presentation of a visual scene to subjects is sufficient for visual illusions to occur. Thus, we infer that perceptual qualities which characterize conscious perception already arise in early stages of memory

3. Experiment 2

In experiment 1, it was shown that the Kanizsa illusion not only influences working memory but also has an impact on iconic memory and fragile VSTM. Therefore, iconic memory and fragile VSTM already contain perceptual qualities such as perceptual inference. To further explore the dependency of the memory capacities on perceptual inference, we manipulated the strength of the illusion, and therefore its visibility. We did this by changing the size of inducing elements (pacmen) and thus the support ratio of the illusory figures, i.e. the ratio of the length of the real inducing contours relative to the total length of the illusory figure. We predicted that when the size of the pacmen increased, memory capacity for the Kanizsa figures would also increase, because the illusion became more obvious, while the capacity of the controls would remain the same regardless of the pacmen’s size. If the larger benefit for better seen illusions would be found, this would confirm the notion that iconic memory and fragile VSTM are stages of memory in which perceptual qualities have arisen.

3.1 Materials and Methods

3.1.1 Subjects

16 out of 20 subjects (11 females and 5 males, aged 19-28, average 23) that participated in the first experiment and passed the training, took part in experiment 2.

3.1.2 Stimuli

Additionally to the figures used in the first experiment, two other sizes of figures were introduced (see Figure 4): bigger figures (2.29° x 3.43°, support ratio = 0.84) and smaller figures (1.71° x 2.86°, support ratio = 0.5). The control conditions were matched and these figures consisted of the same pacman sizes. Thus, there were six sets of stimuli, which constituted an experimental and a control condition for each of three possible sizes.

Figure 4. Three sizes of figures used in experiment 2: small (1.71° x 2.86°), medium (2°x3.15°) and large (2.29° x 3.43°).

3.1.3 Task

The task in Experiment 2 was the same as in Experiment 1.

3.1.4 Procedure

Subjects first performed 12 trials chosen randomly from a total of 36 conditions (support ratio (3) x cue-timing (3) × figures (2) x change/no change (2). They then performed 3 blocks of the task, each divided into four parts of 72 trials. This resulted in 48 trials for each condition (support ratio (3) x cue-timing (3) × figures (2)), resulting in a total of 864 trials.

3.2 Results

In this experiment we predicted a linear increase in performance over size condition for all types of memory for Kanizsa figures and the same level of performance in all size conditions for the control figures. Such a linear increase in performance over size condition for Kanizsa figures but not for the control figures would indicate that the visibility of the illusion can positively affect subjects’ performance. We also expected the effect of the inducers (pacmen) size to be bigger for iconic memory and fragile VSTM than for working memory, due to their bigger capacity.

Repeated measures ANOVA revealed a general significant effect of size (F (2, 14) = 57.931, p < .001). Like in experiment 1 there was a significant effect of memory type: performance decreased over memory conditions (F (2, 14) = 24.675, p < .001) and type of figure: subjects performed better in Kanizsa figure trials than in control trials (F (1, 15) = 74.775, p < .001). We found a significant interaction between size and memory type (F (4, 12) = 6.663, p = .005): subjects’ performance decreased over memory condition for all sizes. There was also an interaction between size and type of figure; for all sizes subjects’ performance was significantly higher for Kanizsa figures than for the controls (F(2, 14) = 3.945, p = .044). However, there was no significant interaction between all three factors: in small size condition differences in capacity between Kanizsa figures and control figures for iconic memory, fragile VSTM and visual working memory were 6.90%, 10.4 % and 5.8% respectively; in medium size condition these differences were 11.7%, 14% and 7.3%, and for working memory: 13.6%, 12.5% and 14.4% (F (4, 12) = 1.082, p = .408).

We found the expected linear increase in performance over size condition for iconic memory and fragile VSTM (see Figure 5A and B) for Kanizsa figures. Differences in capacity of iconic memory between small and medium figures and between medium and big figures were 9.6 % and 4.5 % respectively. For fragile VSTM differences between small and medium figures and between medium and big figures were 6.6 % and 4.1 % respectively. Contrary to our expectations, there was also a linear increase in performance for iconic memory and fragile VSTM in control figures condition (Fig. 5A and B). Nevertheless, it was not as big as the effect of the size

9

of the inducers in Kanizsa figures. Differences in capacity between small and medium figures and between medium and big figures were 4.8 % and 2.6% for iconic memory, and 3 % and 5.7 % for fragile VSTM.

Since in iconic memory and fragile VSTM condition both types of figures benefit from the larger size, we cannot conclude that the strength of the illusory percept is the only explanation of the subjects’ increase in performance. The increase in performance could also be due to an increase in visibility of the pacmen figures themselves. Subjects’ performance also increased over size condition for working memory in Kanizsa illusion trials (see Figure 5C). However, difference in capacity of working memory between small and medium figures was much smaller (0.9%) than difference between medium and big figures (7%). On the other hand, in all size conditions performance on control figures for working memory remained at the same level (see Figure 5C). Thus, in opposition to our predictions, working memory had a largest benefit from the biggest size of the inducers of illusory figures. 50 55 60 65 70 75 80 85 90

small medium big

% C o rrect Figure size

Iconic memory

small medium big

% C o rrect Figure size

Fragile VSTM

A

B

Figure 5. Experiment 2 Results: Percentage correct on the Change Detection task for iconic memory (A), fragile VSTM (B)

and working memory (C) on the Kanizsa figures and the controls for 3 sizes of figures: small, medium and large.

3.3 Discussion of Experiment 2

The results of experiment 1 showed that Kanizsa illusion had a positive effect on subjects’ performance in a change detection task compared to control trials in which the same physical features were present, but no illusory Kanizsa figures could be formed. In this study we manipulated the size of the figures, and therefore the visibility of the illusion, to further explore the dependency of the memory capacities on perceptual inference. We predicted that when the size of the figure increased, memory capacity for the Kanizsa figures would also increase (illusion became more visible), while that of the controls would remain at the same level. A linear increase in performance over size condition for Kanizsa figures but not for the control figures would indicate that the strength of the illusion affects subjects’ performance. We expected this effect to occur not only in working memory but also in iconic memory and fragile VSTM conditions. This would confirm the notion that iconic memory and fragile VSTM are stages of memory in which perceptual qualities have arisen.

However, it was found that iconic memory and fragile VSTM benefit from the larger size of figures in both Kanizsa figures and control trials. Therefore we cannot conclude that the strength of the illusory percept is the only explanation of the subjects’ increase in performance for these types of memory. On the contrary, working memory considerably increased in illusion trials but remained at the same level in control trials. This suggests that working memory benefits most from the strongest and most visible illusory percepts.

One possible interpretation is that iconic memory and fragile VSTM are contrast-dependent forms of memory, which benefit from the amount of white image on a black background, whereas working memory benefits solely from perceptual inference. Nevertheless, for all types of memory, subjects performed better in Kanizsa trials than in control trials, where figures have the same physical properties and the same amount of white image. Therefore, contrast dependence cannot be the only explanation of the increased capacity of iconic memory and fragile VSTM in Kanizsa trials. These results rather suggest that all types of memory had a benefit from perceptual inference, but working memory benefits most. More research e.g. with isoluminant figures is needed to investigate the effect of Kanizsa illusion on iconic memory and fragile VSTM without a contrast.

50 55 60 65 70 75 80 85 90

small medium big

% C orre ct Figure size

Working memory

C

11

4. General Discussion

The heart of this research was the distinction between access consciousness linked to attention and reportability, and phenomenal consciousness (Block, 1990, 2007). Although some researchers tend to associate consciousness with attention, it is suggested in a model by Lamme (2003, 2006, 2010) that attention and consciousness should not be unified. Rather, attention is determined by the depth of processing, whereas consciousness relies on recurrent processing. This model distinguishes four stages of processing: stage 1 and 2 are characterized by a feedforward sweep and consequently unconscious; stage 3 is a recurrent but unattended and thus shallow processing; stage 4 represents deep, recurrent processing spread to the frontal brain areas due to attention. One question that needs to be answer is whether shallow processing of recurrent nature (stage 3) has the same perceptual qualities as deep recurrent processing (stage 4). To investigate this issue, it is crucial to assess whether the representation stored in short term memory has perceptual qualities that characterize conscious perception.

In this research we examined whether the representations stored in fragile short term memory (VSTM) and iconic memory have perceptual qualities i.e. perceptual inference. Previous studies suggest that representations in iconic memory and fragile VSTM are richer and more detailed than information that is accessible and available to report (Landman et al. 2003, Sligte at al. 2008) At the same time, fragile VSTM can be formed independent of attention (Vandenbroucke et al., 2011). This suggests that fragile VSTM and maybe even iconic memory, although unattended, represent information that was consciously perceived. In addition to previous findings, the results of the present studies bring evidence that iconic memory and fragile VSTM are also sensitive to perceptual inference. To manipulate perceptual inference we used change detection task with Kanizsa illusion which requires conscious processing of the spatial context (Harris et al., 2011). We compared the capacity of iconic memory and fragile VSTM with the capacity of visual working memory. In the model proposed by Lamme fragile VSTM incorporates information created in recurrent but unattended and thus shallow processing (stage 3). On the other hand, working memory represents deep, recurrent processing spread to the frontal brain areas due to attention (stage 4). Therefore, the broad question addressed in this research was whether attention is necessary for visual illusion (and thus conscious perception) to occur.

In our experiments we showed that Kanizsa illusion had a positive effect on subjects’ performance in a change detection task compared to control trials in which the same physical features were present, but no illusory Kanizsa figures could be formed. This effect occurred not only for working memory, but also for iconic memory and fragile VSTM. Since iconic memory and fragile VSTM store representations that are formed in stage 3 processing, the results of our experiment demonstrate that objects that are not fully attended, can produce perceptual qualities that characterize conscious percepts. This supports the notion that consciousness and attention are distinct processes.

In experiment 2 we expected that the strength of the illusion would have bigger effect on iconic memory and fragile VSTM than on working memory. Contrary to our predictions, the results suggest that although all types of memory had a benefit from illusion, working memory benefits most from the strongest and most visible illusory percepts. It might be that iconic memory and fragile VSTM are contrast-dependent forms of memory, which benefit from the amount of white image on a black background, whereas working memory benefits solely from perceptual inference. It is also possible that without the contrast Kanizsa illusion would have a smaller impact on iconic memory and fragile VSTM. This would suggest that perceptual qualities have arisen in stage 3 processing (and are stored in iconic memory and fragile VSTM), but attention makes the illusory percept stronger. More research with isoluminant figures is needed to investigate the effect of Kanizsa illusion on iconic memory and fragile VSTM without a contrast.

Another idea for the future research is to investigate the strength of the illusion and manipulate support ratio without changing the size of the inducers. This modification would allow controlling for the effect of contrast and thus providing stronger evidence for the dependency of the memory capacities on perceptual inference.

5. Conclusion

In the present study we showed that illusions could be perceived by subjects even when subjects were presented with the visual scene only for a very brief moment, in a way that prevented from focusing attention on all objects. This would suggest that conscious perception precedes attention. It may turn out that unattended perception has all the properties that matter for consciousness (except for access): as proposed by researchers like Block (1990, 2007), Lamme (2003, 2006, 2010), Koch and Tsuchiya (2007) phenomenal consciousness exceeds access.

References

Block, N. (1990). Consciousness and accessibility. Behavioral and Brain Sciences, 13(4), 596–98.

Block, N. (2007). Consciousness, accessibility and the mesh between psychology and neuroscience. Behavioral and Brain Sciences 30 (5-6), 481–548

Dehaen, S., Changeux, J-P., Naccache, L., Sackur, J., Sergent, C. (2006). Conscious, preconscious, and subliminal processing: a testable taxonomy. Trends in Cognitive Sciences, 10 (5), 204-211

de Gardelle, V., Sackur J., Kouider, S. (2009). Perceptual illusions in brief visual presentation. Consciousness and Cognition, 18(3), 569-577.

Harris, J., Schwarzkopf, D. S., Song, C., Bahrami, B., Rees, G., (2011). Contextual Illusions Reveal the Limit of Unconscious Visual Processing. Psychol Sci., 22 (3), 399-405.

Koch, C., Tsuchiya, N. (2007) Attention and consciousness: two distinct brain processes. Trends in Cognitive Sciences, 11 (1), 16-22.

Kouider, S., de Gardelle, V., Sackur, J., Dupoux, E. (2010). How rich is consciousness? The partial awareness hypothesis. Trends in Cognitive Sciences, 14, 301-307.

Lamme, V. A. (2010). How neuroscience will change our view on consciousness. Cognitive Neuroscience, 1 (3), 204-220.

Lamme, V. A. (2006). Towards a true neural stance on consciousness. Trends in Cognitive Sciences. 10 (11), 494-501.

Lamme, V. A. (2003). Why visual attention and awareness are different. Trends Cognitive Sciences. 7, 12–18. Landman, R., Spekreijse, H., and Lamme, V. A. (2003). Large capacity storage of integrated objects before change blindness. Vis. Res. 43, 149–164.

Merikle, P.M., Joordens, S. (1997) Parallels between perception without attention and perception without awareness. Conscious. Cogn. 6, 219–236.

Posner, M.I. (1994) Attention: the mechanisms of consciousness. Proc. Natl. Acad. Sci. U. S. A. 91, 7398–7403. Sligte, I.G., Vandenbroucke A. R., Scholte, H. S., Lamme, V. A. (2010). Detailed sensory memory, sloppy working memory. Frontiers in Psychology, 1.

Sligte, I.G., Scholte, H. S., Lamme, V. A. (2008). Are There Multiple Visual Short-Term Memory Stores? PLoS One 3(2).

13

Sperling, G. (1960). "The information available in brief visual presentations". Psychological Monographs, 74, 1-29.

Vandenbroucke A. R., H., Sligte, I.G., Lamme, V. A. (2011). Manipulations of attention dissociate fragile visual short-term memory from visual working memory. Neuropsychologia, In press

Weiskrantz, L. (1997) Consciousness lost and found: A neuropsychological exploration. Oxford University Press.