Blindsight-like detection of upright versus inverted
faces under continuous flash suppression in the
absence of subjective awareness
Ellis Emanuel
Bachelor Thesis Brain and Cognition
Universiteit van Amsterdam
Student number: 10583157 Supervisor: Timo Stein
Date: 01-06-2018
Abstract
The capacity for consciously experienced info is limited and therefore it is thought that nonconscious representations of sensory information compete for access to awareness. A popular method to study access to awareness is called breaking continuous flash
suppression (b-CFS). A well-known effect found with b-CFS is the face inversion effect, which shows that upright faces break into awareness quicker than inverted faces. To find out if this effect is present with “blindsight-like” detection (detection with reduced awareness), current study examined if it is possible to obtain above-chance localization, with an
advantage of upright over inverted faces, in the absence of subjective awareness. There was no above-chance localization for faces that were rated as “unseen” and at zero subjective awareness face orientation did not matter. Upright faces were however located more
accurately, consistent with the face inversion effect. Whether this is caused by nonconscious processing remains debatable.
Blindsight-like detection in the absence of subjective awareness
Only a small part of the sensory information from the environment that hits on our retina is consciously experienced (Baars, 1997; Dennett, 2001). The capacity for this sensory information is limited and therefore it is thought that nonconscious representations of sensory information compete for access to awareness (Koch, 2004). Several studies have shown either a change in neural activity when a stimulus is “seen” or no change in neural activity when a stimulus fails to reach consciousness; meaning that it is thought that detection of stimuli is an all-or-none phenomenon (Dehaene & Naccache, 2001; Lamme, Super, Landman, Roelfsema, & Spekreijse, 2000). Multiple theories predict that recurrent interactions between remote brain areas are necessary for conscious perception (Dehaene, Kerszberg, & Changeux, 1998; Dehaene & Naccache, 2001; Di Lollo, Enns, & Rensing, 2000; Lamme & Roelfsema, 2000). One of these theories, for example, is the ‘Global Workspace Model’, which suggests a fleeting memory capacity in which only one piece of content can be dominant at any given moment. This model also suggests that only conscious content is globally distributed to functionally specialized subsystems that are involved in interactions to guide non-automatic goal-oriented behavior (Baars, 1997).A popular method to study access to awareness is called breaking continuous flash suppression (b-CFS; Jiang, Costello, & He, 2007; Stein, Hebart, & Sterzer, 2011a).
Continuous flash suppression (CFS) is a technique in which high-contrast masks flashed to one eye render a stimulus presented to the other eye invisible for a couple of seconds, after which the earlier suppressed stimulus becomes visible (Tsuchiya & Koch, 2005). In the paradigm “breaking continuous flash suppression”, continuous flash suppression is used and
the time it takes observers to detect the suppressed stimulus is used as the measure of access to awareness (Gayet, Van der Stigchel, & Paffen, 2014). In this paradigm, unconscious processing is inferred when (a) detection reaction times to initially invisible stimuli differ, and (b) no comparable differences are found in non-rivalrous control conditions that are supposed to measure non-specific threshold differences between stimuli (Stein, Hebart, & Sterzer, 2011a). Perhaps the most robust and well-replicated effect found with b-CFS is the face inversion effect. This effect shows that the orientation in which faces are shown
influences suppression times; upright faces break into awareness quicker than inverted faces, which is also true for schematic representations of faces (Stein, Peelen & Sterzer, 2011b).
The differences in suppression times in b-CFS can be interpreted as differences in unconscious processing between stimuli; the stimulus with a shorter suppression time acquires access to awareness more quickly. From these differences in visible stimuli, one can infer the differences in the processing of invisible stimuli. Some effects that are already found show that behavior can be influenced before any information has reached conscious awareness. For instance, the bouba kiki effect, which states that there is a tendency for people to match particular linguistic sounds to particular shapes; words like “bouba” are more often matched with a curved shape, whereas words like “kiki” are more often matched with an angular shape (Ramachandran & Hubbard, 2001). In the study of Hung, Styles, and Hsieh (2017) to this effect, a high-contrast dynamic pattern mask was presented to the dominant eye, while the stimulus, the words “bouba” or “kiki” presented in black text inside a white shape (either rounded or angular), was presented to the non-dominant eye. Under this continuous flash suppression, congruent stimuli (e.g., “bouba” inside a curved shape) broke into conscious awareness quicker than incongruent stimuli (e.g. “bouba” in an angular
shape). Results from these studies have been interpreted as showing that the congruence between a word and a shape are processed before either of the two have been consciously perceived. Meaning that the behavior is already influenced when neither the word nor the object has been seen.
Most studies studying access to awareness do not consider objective and subjective measures of awareness; making it difficult to determine if detection is driven by non-conscious versus non-conscious processes (Stein et al., 2011a). The objective measure of
awareness is linked to the forced choice localisation task (where above chance performance is seen as indicative of a level of awareness), while the subjective measure of awareness refers to the level of confidence in noticing the stimulus (Vieira, Wen, Oliver, & Mitchell, 2017). By considering both measures of awareness, one can look at the conscious versus the non-conscious detection of stimuli. The response in b-CFS is to a visible stimulus though (and not to an invisible stimulus). Thus, the interpretation, that the differences in
suppression times in b-CFS can be interpreted as differences in unconscious processing between stimuli, remains just an inference. Nevertheless, researchers interpret the effect as if it is blindsight-like detection (detection with reduced awareness) of an invisible stimulus. This interpretation of the blindsight-like detection of an invisible stimulus has not actually been tested yet, with some exceptions. For example, in the study by Oliver, Mao, and Mitchell (2015); in which is concluded, using an objective and a subjective measure of awareness, that blindsight was found for upright fearful faces but not for inverted fearful faces. Vieira et al. (2017) also measured both localization accuracy and confidence ratings to independently evaluate both the occurrence of vision without awareness and the rate of subjective awareness in their study, in which they examined if emotional salience alone can
drive enhanced subjective awareness and unaware detection. Results of this study support the idea that fear conditioning strength improves both subjective (confidence ratings) and objective awareness (detection of stimulus). More importantly however, is that results suggest that at zero subjective awareness there is still an advantage of conditioned stimuli in localization accuracy, which we refer to as blindsight.
The present study tested whether we can obtain blindsight-like above-chance localization in the absence of subjective awareness under continuous flash suppression. Fixed presentation times were used for the stimuli presented. The stimuli in this study were pictures of upright and inverted faces. Besides testing blindsight-like above-chance
localization, the advantage of upright over inverted faces in breaking CFS, in the absence of subjective awareness, was also tested. A measure of localization performance and a measure of subjective awareness were collected on every trial. For the measure of localization, participants were forced to choose the location of the stimulus; either left or right. Accuracy for the different presentation times was used as a measure, instead of reaction time for example, to be able to also collect visibility ratings.
For the measure of subjective awareness (participants’ confidence ratings) participants were forced to choose from a Perceptual Awareness Scale (PAS). The PAS scale is a four-point scale to assess the degree of visibility of stimuli presented. The PAS scale consists of the categories: “No experience of a stimulus”, “I saw a brief glimpse,” “I had an almost clear experience,” and “I had a clear experience” (Sandberg & Overgaard, 2015).
The hypotheses of the present study are (1) that we can obtain blindsight-like above-chance localization in the absence of subjective awareness under CFS and (2) that, in the
absence of subjective awareness, there is still an advantage of upright over inverted faces in breaking CFS.
Materials and Methods
Participants
Fourteen healthy individuals (13 women; M(age) = 30.4 years, SD 13.47 years) were recruited, at the University of Amsterdam. All participants were in good health and had normal or corrected-to-normal vision. All participants granted informed consent, no
compensation was given for their participation. This study was approved by the University’s Health Sciences Research Ethics Board.
Statistical power
The calculated power for an effect size of d = 0.5, with the number of participants that took part in current study (n = 14), is 0.41.
Stimuli and task
Continuous flash suppression (Tsuchiya & Koch, 2005) was used to suppress awareness of the stimuli. CFS includes dichoptic presentations that induce interocular rivalry, with one image being suppressed by the other. The stimuli, upright or inverted faces, were presented to the dominant eye. The masks, flashing Mondrian-like images were presented to the non-dominant eye. Stimuli were neutral faces from the Karolinska Directed Emotional Faces (Lundqvist, Flykt, & Öhman, 1998), that were converted to greyscale. The Mondrian likeflashing images, consisting of small randomly placed white, black and grey
circles of varying sizes, were created in Matlab (The MathWorks, Natick, MA, USA). Rectangular frames enclosed the Mondrian images as well as the frame for the stimulus presentation. These frames measured 284 x 284 pixels, with noise borders of 8 pixels wide. The stimulus area (rectangle around the face) measured 72 x 72 pixels and this area was 42 pixels away from fixation point (either left or right). All stimuli were presented on a
consistent medium grey background. The task was displayed on a Dell monitor (resolution 1920 x 1080, refresh rate 100 Hz). Observers looked at a pair of dichoptic displays through a custom-built mirror stereoscope, with their head stabilized by a chin rest to guarantee a viewing distance of 50 cm and proper convergence. Visual stimuli were presented with Matlab Psychtoolbox.
Procedure
All trials started with a fixation cross presented to both eyes for 1000 ms, followed by no fixation cross for 400 ms, after which stimulation started. Presentation times differed between 200 ms, 400 ms and 800 ms, with a flash of the mask every 100 ms. In every presentation time 3 masks were presented to both eyes. These are also presented to the dominant eye to prevent afterimages; an impression of the image retained by the eye after stimulation has ended. Over the first 200 ms the face is faded in, slowly increasing in contrast. Ocular dominance was evaluated beforehand, using the hole-in-card test as a modified version of the original ABC test (Miles, 1930). After that, 96 trials were presented where participants only had to locate the face (no confidence rating), to test whether
accuracy was below 70% with the determined dominant eye; meaning suppression is ensured. Then for the main experiment, participants were instructed to maintain fixation on
the fixation cross when detecting the face. When no face was detected, participants were told to go with their “gut feeling”. After presentation time had ended, participants were asked to locate the face by either pressing the left or the right arrow key, providing a measure of localization performance. After that, they had to rate their confidence in their response on the 4-point PAS scale (1 = “No experience of a stimulus”, 2 = “I saw a brief glimpse,” 3 = “I had an almost clear experience,” and 4 = “I had a clear experience”) with the corresponding number key. The procedure is shown in Figure 1. In other breaking CFS paradigms, the time to locate a stimulus is assessed. The current study differs from those paradigms, since both a measure of localization performance and confidence ratings are independently assessed. Current study consisted of 384 trials in total with 16 different faces, each presented upright and inverted, both to the left and the right side of the fixation cross for 3 different
presentation times. All stimuli were presented twice in random order.
A significance level of α = 0.05 is used for all analyses. The analyses are performed in IBM SPSS Statistics 24 and MATLAB 2017.
Non-dominant eye Dominant eye 1000ms 400ms nofixationcross 200,400or800ms 1 2 Confidence 1 – No experience 2 – A brief glimpse 3 – Almost clear exp. 4 – Clear experience Notimelimit 384 trials; 16 faces à upright and inverted; both left and right for 3 pres. times (all shown twice)
Figure 1. The procedure of a trial. A face was gradually introduced to the participant’s dominant eye, while high-contrast masks were presented to the non-dominant eye using continuous flash suppression (CFS). Participants localized as accurately as possible if the face appeared either left or right of the fixation cross, after which they had to rate the confidence in their response. Faces were presented either upright (as shown here) or inverted.
Results
We first analysed the overall accuracy in face detection of all participants with a 2x3 ANOVA design, with orientation (either upright or inverted) and the three different
presentation times. There was a significant main effect of face orientation, F(1,13) = 10.45, p = 0.007, ηp² = 0.45, with higher accuracy for upright faces than for inverted faces. There also was a significant effect of presentation time, F(2,26) = 5.55, p = 0.010, ηp² = 0.30, with higher accuracy for longer presentation times. There was no significant interaction effect
between face orientation and presentation time, F(2,26) = 0.53, p = 0.593, ηp² = 0.04 (see Figure 2).
Figure 2. The detection accuracy for the three presentation times for upright and inverted faces.
Second, we analysed the mean subjective awareness in face detection of all
participants with an ANOVA equal to the first design used. There was a significant effect of face orientation on subjective awareness, F(1,13) = 5.90, p = 0.030, ηp² = 0.31, as well as presentation time, F(2,26) = 5.66, p = 0.009, ηp² = 0.30. In addition, there was no significant interaction effect between face orientation and presentation time, F(2,26) = 3.12, p = 0.061, ηp² = 0.19 (see Figure 3). So, besides the fact that upright faces are seen better and easier to localize than inverted faces, this orientation effect and the time a face is presented also influences the subjective awareness.
Figure 3. The subjective awareness rating for the three presentation times for upright and
inverted faces.
Third, to find out if we can obtain blindsight-like above-chance localization in the absence of subjective awareness, we performed a one-sample t-test. We analyzed the accuracy, depending on subjective awareness, against chance level, for both upright and inverted faces. There was no significant effect for upright faces nor for inverted faces that were rated as “unseen”, t(13) = 1.09, p = 0.294, d = 0.292.
Finally, to see if there is still an advantage of upright over inverted faces in the absence of subjective awareness, we performed a paired samples t-test. We analyzed the accuracy depending on subjective awareness, for upright versus inverted faces. There was no significant effect of “unseen” upright faces against “unseen” inverted faces, t(13) = 1.68,
In Figure 4 the results of the confidence ratings are shown; the proportion of 1-4 presses on the PAS scale on different presentation times for upright and inverted faces.
Figure 4. The confidence ratings; the proportion of 1-4 presses on the PAS scale on different
presentation times for upright and inverted faces is shown.
Discussion
The current study examined if blindsight-like above-chance localization can be obtained in the absence of subjective awareness under continuous flash suppression. There was no effect for upright and inverted faces that were rated as “unseen”; subjects were unable to localize ‘invisible’ faces at above-chance level. Besides, there was no effect for upright faces compared to inverted faces in the absence of subjective awareness. Meaning that at zero subjective awareness it did not matter if faces were shown upright or inverted.
compared to inverted faces, which is consistent with the earlier found face inversion effect (Stein et al., 2011b). Upright faces were furthermore rated with higher subjective awareness compared to inverted faces. Finally, there was an effect found of presentation time; the longer a face was presented, the higher the localisation accuracy and the higher the subjective awareness rating.
It was hypothesized that there would be blindsight-like above-chance localization in
the absence of subjective awareness, with an advantage of upright over inverted faces. Nevertheless, results are not in line with the expectations. An explanation for these unexpected results could be that the sample size was too small. This small sample size resulted in a power of 0.41, meaning that the probability to detect an effect, assuming that there was an effect, was very low. The study by Vieira et al. (2017), in which above-chance localization was found in the absence of subjective awareness, included fifty-two
participants, to ensure a power of 0.95. It is therefore recommended to include more participants in future research.
Then, it could also be that results were influenced by the setting and the length (approximately 45 minutes) of the experiment. The mirror stereoscope and the chin-rest are new and rendered participants ill at ease, possibly influencing their focus. Also, it is possible that participants got tired over time, with only 3 breaks of 10 seconds, which could have influenced their focus as well. Further, due to the instability of both the stereoscope and the chin-rest, it is possible that the exact perceptual field differed a tiny bit between every participant. Some participants even complained about getting dizzy. This could be caused by a non-proper convergence, which perhaps influenced the participants’ focus as well. A more solid construction and longer breaks could prevent this from happening in future research.
A third explanation could be that there is a difference in gender in task performance. Thirteen of the fourteen participants in current study were women. It seems unlikely though that this caused these outcomes, as women often perform better on tasks demanding object location and recognition (Li, 2014) and on face recognition in general (Herlitz, Nilsson, and
Bäckman, 1997). In future research, it might be interesting to look at these gender
differences, as well as to take into account both men and women to make the outcomes more generalizable.
A final explanation could lie in the unfamiliarity with the photographs of the faces. The photographs did not look like everyday faces, probably caused by the fact that they were shown without hair. O’Donnell and Bruce (2001) have indeed shown that hair in particular is an important characteristic for recognizing unfamiliar faces. This effect was also found by Wright and Sladden (2003); participants in this study said they remembered a face between 20% and 25% of the time when originally shown with hair, whereas this remembering rate was only 2% for faces shown without hair. For future research, it is therefore recommended that photographs are used in which the hair is visible. It is also proposed to use pictures of “famous” people. Since there was above-chance localization found (with zero subjective awareness) in the study by Vieira et al. (2017), in which conditioned stimuli were used.
Current study provided evidence for the face inversion effect; upright faces are located more accurate, compared to inverted faces. Stein, Reeder and Peelen (2016) did a study to awareness for faces and objects of expertise. They found that greater car expertise is associated with larger car inversion effects in b-CFS and that there is a similar association between identification performance and inversion effects for faces. Thus, indicating a shared
mechanism governing access to visual awareness, possibly shaped by expertise. These results demonstrate that access to awareness under CFS perhaps does not need to involve innate neural mechanisms, which were initially thought to be a possible cause of the face inversion effect (Johnson, 2005), and that these objects of expertise and faces, can be enhanced before becoming available for conscious access. Stein, Sterzer, and Peelen (2012) also imply that the face inversion effect occurs when a non-consciously extracted structure conforms to an upright human face (causing a stimulus to be detected more quickly). Although the results of these studies suggest that there is some form of nonconscious visual processing, present study found that upright faces were rated with higher subjective
awareness compared to inverted faces; indicating that upright faces are consciously perceived faster than inverted faces. This leads to believe that the face inversion effect is caused by something else rather than unconscious processing differences. The biggest support for this believe arises from the fact that no effect was found in current study for upright faces compared to inverted faces in the absence of subjective awareness; showing that the face inversion effect decreases without consciousness. Considering the fact that expertise also provides an inversion effect, habituation seems more likely to be the cause of upright representations breaking into awareness quicker than an innate mechanism.
As yet we can conclude that we cannot obtain above-chance localization of faces without subjective awareness. This does not rule out the existence of blindsight-like detection or unconscious processing of visual stimuli. Perhaps with the previous done recommendations it is possible to find blindsight-like above-chance localization at zero subjective awareness. Additional research is necessary though, to better understand blindsight-like detection, its causes and its implications.
Literature
Baars, B. J. (1997). In the theater of consciousness: The workspace of the mind. Oxford, United Kingdom: Oxford University Press. http://dx.doi
.org/10.1093/acprof:oso/9780195102659.001.1
Dehaene, S., Kerszberg, M., & Changeux, J. P. (1998). A neuronal model of a global workspace in effortful cognitive tasks. Proceedings of the national Academy of
Sciences, 95(24), 14529-14534.
Dehaene, S., & Naccache, L. (2001). Towards a cognitive neuroscience of
consciousness: basic evidence and a workspace framework. Cognition, 79(1-2), 1-37.
Dennett, D. (2001). Are we explaining consciousness yet?. Cognition, 79(1-2), 221-237.
Di Lollo, V., Enns, J. T., & Rensink, R. A. (2000). Competition for consciousness among visual events: the psychophysics of reentrant visual processes. Journal of
Experimental Psychology: General, 129(4), 481.
Gayet, S., Van der Stigchel, S., & Paffen, C. L. (2014). Breaking continuous flash suppression: competing for consciousness on the pre-semantic battlefield. Frontiers
Herlitz, A., Nilsson, L. G., & Bäckman, L. (1997). Gender differences in episodic memory. Memory & cognition, 25(6), 801-811.
Hung, S. M., Styles, S. J., & Hsieh, P. J. (2017). Can a word sound like a shape before you have seen it? Sound-shape mapping prior to conscious awareness.
Psychological science, 28(3), 263-275.
Jiang, Y., Costello, P., & He, S. (2007). Processing of invisible stimuli: Advantage of upright faces and recognizable words in overcoming interocular suppression.
Psychological science, 18(4), 349-355.
Johnson, M. H. (2005). Subcortical face processing. Nature Reviews Neuroscience,
6(10), 766.
Koch, C. (2004). The quest for consciousness: A neurobiological approach. Greenwood Village, CO: Roberts & Company Publishers.
Lamme, V. A., & Roelfsema, P. R. (2000). The distinct modes of vision offered by feedforward and recurrent processing. Trends in neurosciences, 23(11), 571-579.
Lamme, V. A., Super, H., Landman, R., Roelfsema, P. R., & Spekreijse, H. (2000). The role of primary visual cortex (V1) in visual awareness. Vision Res, 40(10-12), 1507-1521.
Li, R. (2014). Why women see differently from the way men see? A review of sex differences in cognition and sports. Journal of sport and health science, 3(3), 155-162.
Lundqvist, D., Flykt, A., & Öhman, A. (1998). The Karolinska directed emotional faces (KDEF). CD ROM from Department of Clinical Neuroscience, Psychology
section, Karolinska Institutet, (1998).
Miles, W. R. (1930). Ocular dominance in human adults. The journal of general
psychology, 3(3), 412-430.
O'Donnell, C., & Bruce, V. (2001). Familiarisation with faces selectively enhances sensitivity to changes made to the eyes. Perception, 30(6), 755-764.
Oliver, L. D., Mao, A., & Mitchell, D. G. (2015). “Blindsight” and subjective awareness of fearful faces: Inversion reverses the deficits in fear perception
Ramachandran, V. S., & Hubbard, E. M. (2001). Synaesthesia--a window into perception, thought and language. Journal of consciousness studies, 8(12), 3-34.
Sandberg, K., & Overgaard, M. (Eds.). (2015). Using the perceptual awareness scale (PAS). Behavioral methods in consciousness research (pp. 181-195). Oxford
University Press, USA.
Stein, T., Hebart, M. N., & Sterzer, P. (2011a). Breaking continuous flash suppression: a new measure of unconscious processing during interocular suppression?. Frontiers in human neuroscience, 5, 167).
Stein, T., Peelen, M. V., & Sterzer, P. (2011b). Adults' awareness of faces follows newborns' looking preferences. PLoS One, 6(12), e29361.
Stein, T., Reeder, R. R., & Peelen, M. V. (2016). Privileged access to awareness for faces and objects of expertise. Journal of Experimental Psychology: Human
Perception and Performance, 42(6), 788.
Stein, T., Sterzer, P., & Peelen, M. V. (2012). Privileged detection of conspecifics: Evidence from inversion effects during continuous flash
Tsuchiya, N., & Koch, C. (2005). Continuous flash suppression reduces negative afterimages. Nature neuroscience, 8(8), 1096.
Vieira, J. B., Wen, S., Oliver, L. D., & Mitchell, D. G. (2017). Enhanced conscious processing and blindsight-like detection of fear-conditioned stimuli under continuous flash suppression. Experimental brain research, 235(11), 3333-3344.
Wright, D. B., & Sladden, B. (2003). An own gender bias and the importance of hair in face recognition. Acta psychologica, 114(1), 101-114.
