Research project
Student: Doris E. Dijksterhuis Student number: 10354409 Credits for project: 26 EC
Period of research project: 06/02/2017 – 16/06/2017 Final date: 06/16/2017
Supervisor: Matthew Self
UvA representative/Co-assesor: Yaïr Pinto
Research institute: Netherlands Institute of Neuroscience Education: Research Master Brain and Cognitive Sciences Track: Cognitive Science
Backward Metacontrast Masking and
the Role of Border-Ownership
Doris E. Dijksterhuis, BSc
Netherlands Institute for Neuroscience
Visual masking is often used as a tool to investigate visual perception. Understanding how visual stimuli are perceived and what makes them visible is one of the big quests in visual perception studies. When a video frame with a central bar is presented briefly and is followed by a video frame with two flanking bars, the central bar becomes less visible. Macknik & Livingstone explained this backward metacontrast masking effect by
suggesting that the transient signals provoked by the central bar are interrupted by the signals of the flanking bars, which in turn decreases the visibility of the central bar. However, several studies seem to contradict this claim, and we propose that the effect of backward metacontrast masking can be explained by the role of border-ownership. Our results show that the masking effect significantly increases as border-ownership changes, compared to unchanged ownership. These findings support the claim that border-ownership plays an important role in backward metacontrast masking.
GENERAL INTRODUCTION
Many scientists study visual perception to try and understand how we perceive visual stimuli in daylight and what the links are with different forms of neural activity. We perceive visual stimuli seemingly effortless, but it takes a tremendous amount of processing and computational actions before visual information is identified by the brain. Different theories about how we become aware of visual stimuli have been proposed. For example, some studies suggest that recurrent processing of a visual stimulus between higher and lower brain areas,
mediated by feedback and horizontal neural connections, is necessary for visual perception (Lamme, 2000). Others propose that visual perception can only occur if it is mediated by a global brain-scale pattern of activity (Sergent & Dehaene, 2004). To study visual perception, it is
convenient to use a stimulus that can be manipulated in its visibility without changing the stimulus itself, to compare the differences. Visual masking can be used as a tool to study visual perception.
In visual masking paradigms a brief stimulus is visible when presented alone (the target), but less visible when presented together with another stimulus (the mask) being temporally and spatially close (Sergent & Dehaene, 2004; Bouma, 1970; Averbach & Coriell, 1961). The advantage of working with visual masking is that, if you want to compare a condition in which a stimulus is perceived as visible with a condition in which the stimulus is perceived as less visible, the same stimulus can be used. The difference in visibility can be obtained by manipulating the mask.
Masks can be used in different ways. For example, a mask can precede or follow a
stimulus (the target), which is called forward-masking and backward-masking respectively (Enns & Di Lollo, 2000). Backward-masking in particular is of interest for theories of visual
perception, because even though the mask appears after a target, it can still decrease its visibility. When a mask spatially overlaps the target it is called ‘pattern masking’, while a mask that is adjacent to the target but nonoverlapping is referred to as ‘metacontrast masking’. In the current study, we will be focusing on backward metacontrast masking.
The mask in a metacontrast masking paradigm does not spatially overlap with the target, but the contours are closely adjacent. Important here is the time interval between the presentation of the target and the mask (Stimulus Onset Asynchrony or SOA). Extremely short or very long SOAs do not make the target less visible, while an intermediate SOA does (Enns & Di Lollo, 2000). A classic explanation for this effect is the ‘two-channel’ theory (Breitmeyer et al., 2004; Ogmen et al., 2003; Enns & Di Lollo, 2000). This theory suggests that a visual stimulus initiates neural activity in two channels. The first channel is a fast but short-lived signal and consists of the transient onset and offset signal of the stimulus. The second channel is a slow but long-lived signal and processes features of the stimulus. The two-channel theory suggest that the fast-acting signals of a backward mask catches up with the sustained activity of an earlier presented target, interrupting it, and thereby causing less visibility of that target.
A more recent explanation for the effect of backward metacontrast masking has been proposed by Macknik & Livingstone (1998). They compared visual masking in humans with electrophysiological measurements in the primary visual cortex of monkeys. The stimulus parameters of the mask that caused decreased visibility of the target in humans also caused inhibition of the transient signals of the target in the primary visual cortical neurons of the monkeys. More specific: a forward mask affected the onset signal and a backward mask affected the offset signal. This seems to suggest that a mask inhibits the transient responses of the target, causing less visibility. Based on these results, Macknik & Livingstone (1998) propose that the interrupted transient signals of a target contributes to the masking effect. Besides the fact that this suggestion implies the counterintuitive idea that the offset signal contributes to the visibility of a stimulus, results of other similar studies are not always compatible with these results. Several studies were able to measure the early transient responses to undetected masked stimuli, some studies even measured them in higher visual areas, suggesting that the mask did not
interrupt the transient signals (Thomson & Schall, 1999; Jeffrey & Musselwhite, 1997; Kovacs et al., 1995; Rolls & Tovee, 1994). This discrepancy in results led us to contrive an alternative explanation for the effect of backward metacontrast masking. In the next section, we will explain the alternative explanation and how it involves border-ownership cells.
Alternative explanation involving border-ownership cells
Border-ownership cells are neurons that encode the side to which a border belongs, based on local contrast information of visual elements (Zhou et al., 2000). These cells assign the border to the figural region and not to the surrounding background region. For example, consider the border within the circle in Figure 1A and B. Figure 1A is likely to be perceived as a white square against a black background rather than a hole in a black surface, because the white square
“owns” the border. In the case of Figure 1B, the border is owned by the overlaying figure. Figure 1C shows the importance of border-ownership assignment in understanding visual scenes: the ink blot makes it possible for us to see the blue B’s in the background, rather than just seeing blue regions. The occluding borders define the figure of the foreground and by correct assignment of the borders in the background, border-ownership enables us to understand the entire visual scene. Zhou et al. (2000) showed that the assignment of borders to figures is reflected in neural activity, and this activity shows
that a border-ownership cell has a significant different firing rate depending on the side of the figure to which the border belongs. This seems to suggest that border-ownership cells have a preference for a certain side of the figure. It is this property of border-ownership cells that we suggest has a role in visual masking.
Consider the target and mask type 1 in the visual masking paradigm depicted in Figure 2. The targets are black bars on a dark grey background. The black bars “own” the borders and will therefore be perceived as a figure on a grey background. The inner borders of the mask that follows the target are at the same location as the borders of the target. These borders now belong to a figure that is on the other side of the border. In this situation, we expect that a flip in border-ownership occurs which causes a change in neural activity in the border-border-ownership cells, as the side of the figure affects their activity (Zhou et al., 2000). We suggest that this flip in border-ownership, and thus a change in neural activity, causes a higher masking effect meaning a decreased visibility of the target.
Figure 1: A. White figure is seen against a dark background (from Zhou et al., 2000). B.
Border within the circle is owned by the overlaying white figure (from Zhou et al., 2000). C. Bregman ink blotted Bs. The blue regions are the same in the left and right figure, but the ink blot owns some of the same borders, making it easier to perceive the Bs (from Williford & von der Heydt, 2014).
The role of border-ownership cells in backward metacontrast masking has not been established yet by previous studies. Given the fact that the proposal of Macknik & Livingstone (1998) is counterintuitive and not consistent with other studies, we hope that our results can provide an alternative explanation.
EXPERIMENT 1
To test whether the flipping of border-ownership indeed facilitates the masking effect, we designed a masking paradigm with two conditions: one with a flip in border-ownership and one with no flip. This was achieved by designing two mask types (see Figure 2). Mask type 1 is described in the section above and mediates the flipping condition. Mask type 2 does not cause a flip in border-ownership, as the border now belongs to the grey ‘ I ’ shaped figure, created between the two big black masks that now belong to the background. The ‘ I ’ shaped figure is on the same side of the border as the target was, so will cause less change in neural activity in border-ownership cells than in the other condition. The local contrast changes on the vertical boundaries of the target are the same for both mask types, meaning that a flip in border-
ownership is only due to the change of the side of the figure where the border is assigned to. We expect to find a greater masking effect, so less visibility of the target, in the flipping condition than in the no-flipping condition. The task that is used in this experiment was to identify the longer bar (see the target in Figure 2: the right bar is longer than the left bar). We expect that a decrease in visibility translates in a decrease in correct responses.
Previous studies show that with increasing duration of a stimulus, the detection performance also increases (Rolke & Hofmann, 2007; Bachmann & Allik, 1976). We expect to see this effect in both conditions, as the target will be presented for different amount of times but in the flipping condition we expect the proportion of correct responses to be smaller than in the non-flipping condition.
Methods
Participants. All eight subjects were employees or interns from the Netherlands Institute for
Neuroscience. The ratio man/woman was 1:1 and the age differed from 21 to 39 years old. Each subject was righthanded.
Stimuli and apparatus. Stimuli in the psychophysical experiment were presented on a Trinitron
T5400 monitor at a refresh rate of 60 Hz in a dimly lit room. The size of the monitor was 47.6° by 39.2° with a chinrest at a viewing distance of 35 cm. During the entire experiment a black fixation cross with 0.6° sized arms was presented at a background of 33.75 cd.m-2 (candelas per metre square luminance) in the middle of the screen (Figure 2). The stimuli consisted of a target
Figure 2: Backward metacontrast masking paradigm. First a fixation cross on a grey
background is depict. Hereafter is the target presented for a duration of 33, 50, 67, 83, 100 or 133 ms, followed by mask type 1 or 2, representing the flipping and non-flipping condition respectively. Participants had to indicate which target was longer (left or right bar) after the mask has disappeared. At the bottom of the figure is the timeline of the paradigm described.
which was either 0.8° by 7.4° or 0.8° by 8.6° and mask type 1 that existed of two bars of 0.8° by 9.7° with a space of 0.8° between them and mask type 2, which was shaped as is shown in Figure 2. In this case the size of the longer part was 0.8° by 11.6° and the size of the shorter part 0.8° by 9.7° with again a space of 0.8° in between them. Targets and masks were always black,
presented in the center of a dark grey square of 3.9° by 11.6° with 8.55 cd.m-2. Everything was programmed in MATLAB R2006.
Procedure and design. Each participant first performed a training session in which they learned
the task. First the target was shown so the participants knew what to search for. The target was presented 6.0° left and right from the fixation cross (Figure 2). One of the targets was always longer than the other, randomly varying between the left and right side. The task was a 2
alternative forced choice (2 AFC) task where the participants had to report, per trial, if the longer target was on the left or the right side of the cross. The duration of the targets could differ per trial between: 33, 50, 67, 83, 100 and 133 ms. The target was always followed by mask type 1 or mask type 2, varying randomly. The duration of the mask was always 300 ms, whereafter the mask disappeared and only the grey square remained. In the training session, the participants received feedback after their response (green dot or a red dot meaning correct or incorrect respectively). If the participants scored only green dots for multiple trials in a row, they could start the real experiment. The real experiment started with an introduction and again an example of the target. After this the experiment started. Feedback was no longer given and a new trial started after the participant gave a response. Pressing the left arrow key meant that they thought the longer target was on the left side of the fixation cross, and pressing the right arrow key meant the same for the right side of the fixation cross. After every 24 trials, a black screen appeared for 4 seconds, which was meant for the participants to rest their eyes and reduce adaptation effects. Hereafter, the background of 33.75 cd.m-2 appeared again, which indicated that the participants had 2 second before the next trial started again. Each participant did 960 trials (40 trials per condition) which took approximately 30 minutes.
Analysis. We used a psychometric function to relate the probability of correct responding to the
duration of the target (Strasburger, 2001). A logistic psychometric function was fitted per person with the Palamedes toolbox for MATLAB, to calculate the threshold and slope per person and per mask. Mask type 1 and mask type 2 correspond to the flipping and non-flipping condition respectively. The difference between the two mask types were tested with a dependent t-test by
using the thresholds of all participants and a separate t-test was done with the slopes of all participants. The statistical analysis was also done in MATLAB.
Results
The percentage correct responses was measured. One participant was excluded from the analysis, due to divergent results, which indicated that he or she did not understand the task well enough. The threshold and slope are calculated with a psychometric function of which one example is shown in Figure 3. Where the function intercepts 0.75 on the y-axis, the
corresponding x-value will be the threshold. The steepness of the function on that point is the
Figure 3: The total proportion correct responses per condition is shown for one subject.
On the x-axis is the target duration in ms depict and on the y-axis is the proportion correct depict. The red and blue line represent the non-flipping (mask type 2) and flipping (mask type 1) condition respectively. The dots represent the goodness-of-fit of the logistic psychometric function. A target duration of 117 ms did not exist in the paradigm, but was implemented here to depict the curve of the psychometric function.
slope. The difference between the thresholds of all participants between mask type 1 (M = 69.1,
SE = 8.2) and mask type 2 (M = 80.1, SE = 9.4) was significant (t (6) = -3.08, p < 0.05, r = 0.78,
Table 1). The difference between the slopes of all participants between mask type 1 (M = 1.42,
SE = 0.24) and mask type 2 (M = 1.44, SE = 0.43) was not significant (Table 2). Excluding the
Figure 4: The threshold per participant and condition is shown. The y-axis represents the
detection threshold in ms. The conditions are depict on the x-axis. The grey bars indicate the mean detection threshold per condition. The red and blue dots represent the mean detection threshold per subject. *** Indicates that the difference between the two means is significant.
Figure 5: The slope per participant and condition is shown. The y-axis represents the slope
of the psychometric function. The conditions are depict on the x-axis. The grey bars indicate the mean slope per condition. The red and blue dots represents the mean slope per subject.
Discussion of Experiment 1
We hypothesized that border-ownership plays a role in backward metacontrast masking. We therefore expected that mask type 1 would cause a flip in border-ownership leading to a decrease in visibility. Compared to the non-flipping condition, this would cause lower
proportions of correct responses of the participants. This in turn should result in lower contrast detection thresholds. As we can see in Table 1 and Figure 4, the thresholds are indeed
significantly lower in the flipping condition than in the non-flipping condition. The difference in the slopes between the conditions on the other hand is not significant (Table 2, Figure 5).
From the psychometric function parameters we used only the threshold and slope values for the statistic tests, as these explain the underlying sensory mechanism the most (Kingdom & Prins, 2010). In the current study the threshold indicates the amount of milliseconds the target had to be presented on the screen, for the participant to reach our criterion of 75% correct responses. Therefore, the difference between the thresholds of the flipping and non-flipping condition was expected to be significant, which it turned out to be (Table 1, Figure 4). The slope on the other hand provides information about the increment of proportion correct per increase in duration which represents the level of noise in the response (Strasbuger, 2001). The change in threshold was not accompanied by a significant change in slope (Table 2, Figure 5).
Our results suggest that border-ownership plays a role in visual masking, because we see that the masking effect changes when only the border-ownership changes and other factors remain nearly constant. However, one factor that did change between conditions was the average luminance of the masks. Mask type 2 had more black surface than mask type 1. This could also have affected the masking effect and we therefore designed a second experiment in which the mask was held constant to control for this issue.
EXPERIMENT 2
The results of experiment 1 show that a flip in border-ownership indeed facilitates masking. It is interesting to investigate if this effect is truly mediated by border-ownership flipping or because of the differences between the masks. As a control we did a second experiment in which we tested this (Figure 6).
The Gestalt laws state that convex object are more likely to be perceived as objects against a background than concave objects are (Wertheimer, 1923). In Figure 6A the target is a single concave object, most likely to be perceived as a figure against a background as it is presented in isolation. When the mask appears, the former target is now surrounded by convex objects which, according to Gestalt laws, are now perceived as the figures and the concave objects as the background. Whatever happens with the contrast, the convex object will always be seen as a figure against a background. If the target was already a convex object, it will still be perceived as a figure when the mask appears (Figure 6B and Figure 7, no flip condition). The situations described above form the flipping and non-flipping condition: a concave target
changes into a background (flipping condition) or a convex target remains a figure (non-flipping condition). In this experiment, a second condition was introduced. In Figure 7 are example trials displayed that show conditions in which the contrast did or did not switch between the target and the mask. We suggest that border-ownership cells play a role in backward metacontrast masking
Figure 6: Paradigm of experiment 2. A. This is an example trial with a concave target.
The target changes to the background when the mask appears (see red square), as the convex objects are more likely perceived as figures and the concave objects as
background. B. This is an example of a convex target. The target stays a figure when the mask appears.
and we know that these cells respond to a change in local contrast information (Zhou et al., 2000), therefore we expect that in the contrast switching situation, the difference between the flipping and non-flipping condition is higher than in the no contrast switch condition.
Methods
Participants. All five subjects were employees or interns from the Netherlands Institute for
Neuroscience. The ratio man/woman was 2:5 and the age differed from 21 to 39 years old. Each subject was righthanded.
Stimuli and apparatus. During almost the entire experiment a black fixation cross with 0.6° sized
arms was presented at a background of 40.62 cd.m-2 in the middle of the screen (Figure 6A). The mask existed of two horizontal strips across the whole screen, 7° above and below the fixation cross, filled with 7 or 8 convex figures which can be black or white (Figure 7). The space
between the convex figures are concave objects which are seen as a background and they always have the opposite contrast of the figures. The height of the strip is 4.7° and the width of one
Figure 7: These are the conditions of experiment 2. In the no flip condition, the target is
a concave object which will be perceived as the background when the mask appears. In the flip condition is the target a convex object, that stays a figure when the mask appears. In the no contrast switch condition, there is no change in contrast between the target and the mask. In the contrast switch condition, the contrast does switch between the target and the mask.
figure is 2.7°. The target consisted of one of these convex or concave figures, randomly switching between black and white across trials. The position of the target could be in the top strip or in the bottom strip, in the middle of the strip or 5.4° to the right or the left of the middle. Subjects had to report whether they saw the target in the top or in the bottom strip. Everything was programmed in MATLAB R2006.
Procedure and design. The procedure was similar to experiment 1. Each participant first
performed a training session in which they received feedback, before they started with the real experiment. In the real experiment the duration of the target varied between: 16, 33, 50, 67, 83, 100 and 133 ms. The target was always followed by a mask (Figure 6). The mask could have the same contrast as the target or the opposite (Figure 7). The duration of the mask was always 300ms, where after the mask disappeared and only the grey background was left. A new session would start after the participant had responded by pressing the up arrow key, indicating they saw the target in the top strip, or the down arrow key, indicating they saw the target in the bottom strip. Each participant did two sessions of 1176 trials. There were 85 trials in total per condition.
Analysis. The analysis was done in the same manner as in experiment 1, only were the flipping
and non-flipping condition calculated per contrast switching of non-contrast switching condition.
Results
This experiment is still in progress. In this paper we will present the preliminary results from the first five subjects. The percentage correct responses was measured. In the contrast switch condition, the difference between the thresholds of all participants between mask type 1 (M = 35.9, SE = 3.8) and mask type 2 (M = 28.7, SE = 5.7) was not significant (Table 3). In the no contrast switch condition, the difference between mask type 1 (M = 35.6, SE = 4.7) and mask type 2 (M = 31.3, SE = 2.8) was not significant as well (Table 3). There were also no significant results in the contrast switch condition between the slopes of all participants between mask type 1 (M = 1.13, SE = .11) and mask type 2 (M = 1.24, SE = 0.12) nor in the no contrast switch condition between mask type 1 (M = 1.40, SE = 0.11) and mask type 2 (M = 1.83, SE = 0.17) (Table 4).
Figure 8: The threshold per participant and condition is shown. The y-axis
represents the detection threshold in ms. The conditions are depict on the x-axis. The grey bars indicate the mean detection threshold per condition. The red and blue dots represent the mean detection threshold per subject.
Figure 9: The slope per participant and condition is shown. The y-axis represents the
slope. The conditions are depict on the x-axis. The grey bars indicate the mean detection threshold per condition. The red and blue dots represent the mean slope per subject.
Discussion of Experiment 2
Again are the thresholds for the flipping condition higher than the non-flipping condition. We also found that the difference between the flipping and non-flipping conditions seems to be bigger in the contrast switch condition than in the no contrast switch condition (Table 3, Figure 8). Despite the fact that the differences were not significant, they still meet our predictions. There was again no significant difference between the slopes (Table 4, Figure 9).
Albeit these results are preliminary, all 5 subjects show higher thresholds, suggesting that an additional 3 subjects would be sufficient to establish significance.
The targets in these experiments varied between convex and concave formed objects. We have not checked if this could have an effect on the detection performance of the participants, therefore this could be a confounding factor.
GENERAL DISCUSSION
The results of experiment 1 together with experiment 2 showed that when the border-ownership flipped between the target and the mask, the masking effect was significantly higher. The difference between the flipping and non-flipping condition was higher when the contrast of the target switched when the mask appeared than when the contrast did not switch. Our results suggest that border-ownership plays a role in backward metacontrast masking. Below follows an explanation of how the observed effect could be explained.
Consider a person looking at the target of experiment 1 being presented on a screen. Projections from border-ownership cells to higher visual areas detect the borders in the image and set up two competing hypotheses: one suggesting that the target is ‘a bar’ and the other suggesting it is ‘a hole’ (Figure 10A). The border-ownership cells that point to the center of the bar or to the surrounding frame of the hole are equally active in the start as they only receive feedforward information about the edge segment in their receptive field. The neural
representation of the ‘bar’ hypothesis will eventually be stronger as it is more likely than the other hypotheses, due to previous experience, as embodied in the Gestalt laws (Wertheimer, 1923). The neurons pointing to the center of the bar win the competition and set up a consistent pattern of border-ownership at lower levels. The cells label the bar region as a figure and due to recurrent processing between different areas, the percept of the bar as a figure will gradually get
stronger. Now imagine the target getting replaced by mask type 1 of experiment 1 (Figure 10B). The border-ownership will initially not change, but there will be a new winning hypothesis at higher levels. The winning high level neurons will now send feedback to the border-ownership cells which will cause a switch in activity. The cells previously pointing to the center of the target, will now be less active and the cells previously pointing to the surrounding area of the target, now point to one of the bars of the mask and will therefore be more active. The perceived figure-ground structure is now switched. If this switch happens fast enough, the target region will not have enough time to become fully established as a figural region and the perception of the target will be impaired. Now imagine that mask type 2 of experiment 1 follows the target (Figure 10C). The new hypothesis that is now created at higher visual areas is consistent with the already activated border-ownership cells, as the cells previously pointing to the center of the target, now point to the center of the ‘ I ’ figure. This means that the figure-ground structure remains the same and therefore, the perception of the target will be relatively unimpaired.
Figure 10: A. After the target is
presented, the borders are detected and hypotheses are made at higher visual areas. Eventually will the ‘bar’ hypothesis win from the ‘hole’ hypothesis. B. After mask type 1 is presented, a new
hypothesis is made, ‘bars’, and this one will win and change the activity of the border-ownership cells. C. After mask type 2 is presented, a new hypothesis is made, ‘I shape’, and this one will win as well, but will not change the activity of the border-ownership cells.
More studies are necessary to further investigate the exact underlying mechanisms of backward metacontrast masking, but we can already suggest that border-ownership plays a significant role in it. As we discussed before, if border-ownership plays a role in masking, this could mean that the transient signals of an undetected masked stimulus get processed beyond the initial processes which indicates that the transient signals play a less important role in masking than suggested by Macknik & Livingstone (1998). Our results are not sufficient to dismiss the importance of the transient signals in visual masking all together, as we only collected
psychophysical data. Performing similar experiments as in the current study while measuring neural activity from the visual field will yield more information about the effect of border-ownership on backward metacontrast masking.
REFERENCES
Averbach, E. & Coriell, A. (1961). Short-term memory in vision. Bell System Technical Journal,
40(1), 309-328.
Bachman, T. & Allik, J. (1976). Integration and interruption in the masking of form by form.
Perception, 5(1), 79-97.
Bouma, H. (1970). Interaction effects in parafoveal letter recognition. Nature, 226(5241), 177-178.
Breitmeyer, B.G., Ro, T. & Ogmen, H. (2004). A comparison of masking by visual and
transcranial magnetic stimulation: implications for the study of conscious and unconscious visual processing. Consciousness and Cognition, 13(4), 829-843.
Enns, J.T & Di Lollo, V. (1997). Object substitution: A new form of masking in unattended visual locations. Psychological science, 8(2), 135-139.
Enns, J.T. & Di Lollo, V. (2000). What’s new in visual masking? Trends in Cognitive Sciences,
4(9), 345-352.
Fahrenfort, J. Scholte, H. & Lamme, V. (2007). Masking disrupts reentrant processing in Human visual cortex. Journal of Cognitive Neuroscience, 19(9), 1488-1497.
Kingdom, F.A.A. & Prins, N. (2010). Psychometric Functions. In Psychophysics. A practical
Kovacs, G., Vogels, R. & Orban, G.A. (1995). Cortical correlate of pattern backward-masking.
Proceedings of the National Academy of Sciences, U.S.A., 92, 5587-5591.
Lamme, V. & Roelfsema, P. (2000). The distinct modes of vision offered by feedforward and recurrent processing. Trends in Neurosciences, 23(11), 571-579.
Lamme, V.A.F. (2000). Neural mechanisms of visual awareness: a linking proposition. Brain
and Mind, 1, 385-406.
Macknik, S.L. & Livingstone, M.S. (1998). Neuronal correlates of visibilitiy and invisibility in the primate visual system. Nature, 1(2), 144-149.
Ogmen, H., Breitmeyer, B. & Melvin, R. (2003). The what and where in visual masking. Vision
Research, 43(12), 1337-1350.
Rolke, B. & Hofmann, P. (2007). Temporal uncertainty degrades perceptual processing.
Psychonomic bulletin & review, 14(3), 522-526.
Rolls, E.T. & Tovee, M.J. (1994). Processing speed in the cerebral-cortex and the
neurophysiology of visual masking. Proceedings of the Royal Society of London, Series B,
Biological Sciences, 257, 9-15.
Sergent, C., Dehaene, S. (2004). Neural processes underlying conscious perception.
Experimental findings and a global neuronal workspace framework. Journal of Physiology, 98, 374-384.
Strasbuger, H. (2001). Converting between measures of slope of the psychometric function.
Perception & Psychophysics, 63(8), 1348-1355.
Thomson, K.G.& Schall, J.D. (1999). The detection of visual signals by macaque frontal eye field during masking. Nature Neuroscience, 2, 283-288.
Wertheimer, M. (1923). Untersuchungen zur lehre von der gestalt II. Psychol Forsch, 4, 301-350.
Williford, J.R. & von der Heydt, R. (2014). Border-ownership coding. Scholarpedia J, 8(10), 30040.
Zhou, H., Friedman, H. & von der Heydt, R. (2000). Coding of border ownership in monkey visual cortex. The Journal of Neuroscience, 20(17), 6594-6611.
