ABSTRACT. When a saccadic eye movement is made toward a visual stimulus, the latency of V1 activity elicited by the stimulus is correlated with saccadic latency [Lee et al., 2010, J. Neurophysiol., 105, 2556-72]. We further examined the changes in V1 activity during the gap saccade task that is known to robustly reduce saccadic latency. Accompanying the gap effect, V1 spike activity elicited by the visual saccadic target systematically changed; neural latency became shorter and firing rate increased, as the gap duration increased. Additionally, before express with respect to regular latency saccades, lower pre-target activity and earlier neural latency and higher firing rate during response epoch were observed. Decision stage downstream to V1 is hypothesized to detect the target appearance based on the temporal contrast of the V1 spike density [Lee et al., 2013. J. Neurosci., 33, 3760-64], and a simulation of temporal contrast of artificial spike trains revealed that the lower level of pre-target activity is systematically related to earlier target detection, in line with the experimental data. These results further demonstrate that the V1 spike activity correlates with the latency of visually-guided saccades.

ABSTRACT. It has been widely shown that emotion interacts with attention. The present study aimed to investigate how emotion influences gaze redirection by applying saccade task. In the study, task-irrelevant emotional (positive, neutral, or negative) faces appeared for 500 ms at the beginning of trials. Later, an endogenous cue pointed to one of six potential target locations which were composed of three possible target locations in each visual field. After 0-400 ms of this initial cue, it was replaced by a new cue pointing toward a new target location in half of trials (inconsistent trials) or stayed unchanged in the remaining trials (consistent trials). Within inconsistent trials, the new target located either in the same visual field as the initial target or in the opposite visual filed of the initial target. Subjects were required to make a saccade to the target as soon as the initial cue appeared and redirect the saccade to the new target in the inconsistent trials. We found that it took longer time to reach the correct target location in inconsistent trials than in consistent trials generally and the presence of emotional faces influenced the saccade performance only in inconsistent trials. When redirecting the saccade to the new target in the same visual field, it took longer time after watching negative faces compared to after watching positive or neutral faces. Additionally, when redirecting the saccade to the new target in the opposite visual field, the saccades landed more precisely after watching positive faces than after watching neutral or negative faces. These results suggest that mere presence of task-irrelevant emotional faces could influence gaze redirection and further different emotion s impact gaze redirection differently.

ABSTRACT. As wearable cameras gains popularity, importance of the first person videos (FPV) processing grows. For many applications, such as object classification, intent classification [2], summarization, retrieval and augmented reality, FPV visual attention model will be useful. Existing datasets are limited to simple scenarios of object interaction where hands and head movements are the main factors [1,2,3].

We recorded eye-tracking data from subjects while they are performing activities (commuting, social interactions, and object manipulations) in actual environments (office, home, public places). Our preliminary data analysis discovers the following: 1. Context-centric. Fixations are determined by the current context. During social interactions, speakers’ faces capture attention. For object manipulation, objects of interests are fixated upon. In walking/running, fixations are divided between direction of travel and ground plane. 2. Multi-tasking. Subjects are frequently engaged in multiple tasks, e.g. walking and talking; object manipulation and walking etc. Their attention is constantly shifting between the tasks. 3. Embodiment. (a) Attention is modulated to avoid hazards and collisions. (b) Gazes are directed towards speaker’s face in social interactions while averted from strangers’ gazes. (c) Oculomotor structure pre-determines the distributions of fixations. There is a strong center bias in fixation distributions. To the best of our knowledge, our dataset is the first publicly available unconstrained FPV with eye-tracking information. There are currently 7 subjects with over 15 hours of data.

References 1. Yamada, K., Sugano, Y., Okabe, T., Sato, Y., Sugimoto, A., & Hiraki, K. (2012). Attention prediction in egocentric video using motion and visual saliency. In Advances in Image and Video Technology 2. Li, Y., Fathi, A., & Rehg, J. M. (2013). Learning to predict gaze in egocentric video. ICCV. 3. Boujut, H., Benois-Pineau, J., & Megret, R. (2012). Fusion of multiple visual cues for visual saliency extraction from wearable camera settings with strong motion. ECCV Workshops.

ABSTRACT. Abstract

In daily life, normal-sighted individuals use vision to assist in tasks such as search, object recognition and the guidance of actions. Additionally, humans use anticipatory eye movements in natural tasks that “look-ahead” to objects that will be used in the near future. Look-ahead fixations (LAFs) are purposeful and influence subsequent visuo-motor coordination of complex everyday actions. However, little is known about the role of such predictive eye movements in other non-human primates. We trained a juvenile 10-year-old female captive-born Sumatran orangutan (Pongo abelii) to wear a portable eye tracker and recorded her gaze in several natural tasks. Here we present data from a 3 dimensional search task where the animal retrieves a reward hidden in one of 24 potential locations (branches) within an array of 6 small model “trees”. Results show that she learned the target location faster when it appeared in a fixed tree and this was accompanied by an increase in anticipatory LAFs to the relevant tree as she learned the location. In a second condition, where the target location was always in the same branch on different trees, she failed to overtly locate the target. Six adult humans also failed to learn this rule in a limited number of trials. Interestingly, there was also an increase in the number of LAFs by the Orangutan to the correct branch despite being unable to locate the target. She had visually learnt the rule that the target would always appear on a certain branch, but could not inhibit motoric reaching behaviour to the incorrect locations. This captive bred Orangutan is capable of making anticipatory eye movements in the same way as humans, and their frequency varies as spatial learning occurs.

ABSTRACT. Saccadic eye movements are often task specific. We use eye movements during routine stereotypical tasks and the sequence of these saccades and fixations reflect the temporal and spatial demands of visuomotor routines. However, once such a stereotypical task has been learnt, it is still unclear how these visuomotor transformations can be transferred to another similar task. Here, we looked at the eye movements of a captive born Sumatran orang utan (n =1) and a number of human participants as they played an iPad game (block stacking) designed for human toddlers. Specifically, we looked for changes in eye movement scanpaths when participants were unexpectedly given a game that differed in configuration to the one they had been trained on. Findings suggest that prior knowledge of the first stacking game is transferred to the new game in both species, but there are subtle differences in fixation duration, fixation number and in the number of anticipatory fixations (look-aheads) between the two different species of hominid that perhaps reflects their differing visuomotor demands. This study not only documents the ability of a captive orang utan to use an iPad in her enrichment program, but it also illustrates how the visual system of these very different primates is driven by the same task demands.

ABSTRACT. There is growing evidence of a common visuotypy related to the degree of autistic tendency, even across the “normal” population, via Baron-Cohen’s Autistic Spectrum Quotient (AQ) [1]. This flies in the face of rather low correlations between the triad of core symptoms of autism in clinical populations. Both electrophysiological and psychophysical evidence will be presented showing that autistic vision is characterized by an afferent magnocellular pathway abnormality – specifically neural recovery after stimulation is impaired as measured by higher amplitude second order Wiener kernels for multi-focal flash VEP. We predicted and found lower flicker fusion frequencies for low temporal contrasts in those with high cf low AQ score. Such an abnormal magnocellular function should also affect saccadic suppression, as saccades have been reported to selectively suppress magnocellularly driven perception [2]. The prediction of a lower degree of suppression in those with low vs high AQ was borne out, with no significant difference in suppression of low (0.3 cpd) and high (2.0 cpd) spatial frequencies for those low in AQ score. With several saccades per second and each episode giving suppression for ~150 ms, it appears that in those with high AQ, high spatial frequencies – or “detail” is being refreshed at the expense of low spatial frequencies. Thus it appears that the saccadic suppression differences with autistic tendency are causal in promoting local over global percept in those with high autistic tendency.

[1] Baron-Cohen, S., Wheelwright, S., Skinner, R., Martin, J. & Clubley, E. 2001 The autism-spectrum quotient (AQ): evidence from Asperger syndrome/high-functioning autism, males and females, scientists and mathematicians. J Autism Dev Disord 31, 5-17. [2] Burr, D.C., Morrone, M.C. & Ross, J. 1994 Selective suppression of the magnocellular visual pathway during saccadic eye movements. Nature 371, 511-513.

ABSTRACT. Clustering is an important unsupervised learning approach, and widely used in computer vision, pattern recognition and machine learning, etc. While a large amount of clustering algorithms have been proposed, existing approaches usually require some a priori knowledge of the data to be clustered in one form or another. Perhaps the most frequent required input parameter is the number of clusters, which is not easy to determine in many cases. While some algorithms are able to determine the number of clusters automatically, they often require other parameters as input. In most cases these input parameters are determined empirically or based on experience. This may be a problem in the application of clustering in practical tasks. With the pairwise similarity matrix of data to be clustered as input, dominant sets clustering has been shown to be an effective data clustering and image segmentation approach, partly due to its ability to find out the underlying data structure and determine the number of clusters automatically. However, we find that the original dominant sets clustering method is sensitive to the similarity measures used in building the similarity matrix. This means that parameter tuning is required to generate satisfactory clustering results, and dominant sets clustering results are also parameter dependent. In order to remove the dependence on parameter tuning, we propose to apply histogram equalization to similarity matrices before they are used in clustering. While this transformation is shown in experiments to remove the sensitiveness to similarity measures effectively, it also results in over-segmentation. We then study some approaches to extend clusters and merge clusters. Both data clustering and image segmentation experiments validate the effectiveness of our dominant sets based clustering algorithms.

ABSTRACT. The purpose of the study was to investigate the relationship between gestalt processing and affect recognition in Autism Spectrum Disorder (ASD). Fourteen ASD and 15 typically developing (TD) male adolescents age 13 to 16 years old participated in the study. A visual Delayed Match-to-Sample memory task involving intact and scrambled scene stimuli was used to measure gestalt and local processing. Affect recognition, which was measured using the Affect Recognition subtest from a neuropsychological test (Neuro-Psychological Assessment-Second Edition), was correlated with their performance on the visual memory task. Additional measures of gestalt and local processing were obtained by administering the Figure Ground and Visual Closure subtests from the Test of Visual Perceptual Skills-Third Edition (TVPS-III). The results showed that participants with ASD had significantly poorer affect recognition than TD participants. The ASD and TD participants did not differ significantly on their performance on the Figure Ground and Visual Closure subtests. However, on the visual working memory task, participants with ASD had lower accuracy scores on the intact stimuli when compared to the TD participants. Both ASD and TD participants had comparable accuracy scores on the scrambled stimuli. Relative to themselves, ASD participants had a comparable performance on intact and scrambled stimuli, while TD participants were significantly better on the intact stimuli than on the scrambled stimuli. This suggested that unlike the ASD participants who had comparable gestalt and local processing, TD participants were better at gestalt processing and this helped with their memory performance. A positive relationship was found between participants’ gestalt processing ability on the visual memory task and their affect recognition. The findings did not fully support the weak central coherence theory. Gestalt processing in ASD may need to be further examined to better understand the conditions under which it affects the ability to recognize emotions in others.

ABSTRACT. Purpose: Orthokeratology has been identified as a method to control myopia progression. The reasons for this positive effect remain elusive but may be related to changes in peripheral refraction during the course of the day. The purpose of this case report is to measure changes in peripheral refraction after orthokeratology lens removal. Methods: Peripheral refraction was measured using an open-field autorefractor in the central 60 degrees for two participants. Both subjects had worn orthokeratology lenses for more than two months. Measurements were taken once in the morning shortly after removal of the orthokeratology lenses and then repeated again in the late afternoon. In addition, topography measurements were performed in one subject. Results: A considerable change in peripheral refraction between the two measurements was observed in both subjects. The change in peripheral refraction was in line with the change of topography. Conclusions: Constant changes of defocus levels in the peripheral retina after orthokeratology might be a possible factor in myopia control. Further analysis in a larger population is needed.

ABSTRACT. Reaction time (RT) is an important measure to evaluate the performance of visual search task. The observers’ responses usually begin a certain latency after a stimulus onset. This latency contains the time taken for a signal transmission and a pattern recognition process. Lleras et al. (2005) reported that RTs faster than the usual latency were observed after resuming interruption of visual search. Enns and Lleras (2008) discussed the rapid resumption (RR) from the interruption could be well explained by an iterative prediction/confirmation mechanism (Fukushima, 1980; Kawato, et al., 1993; Rao & Ballard, 1999); the top-down prediction was compared with a sensory input and then confirmed. Here we examined how the RRs were modulated when mask patterns were presented while interruption. There should arise an inappropriate error signal when a stimulus was replaced with a mask pattern. In this case, the top-down prediction would be distorted, and the visual system could not resume the interrupted search from the confirmation process. The result showed the masking suppressed the RRs from the interruptions, and the degree of suppression depended on mask patterns. This result supports the prediction/confirmation mechanism for the visual search task.

ABSTRACT. Previous research suggests that background noise is beneficial for memory performance of children with Attention-Deficit / Hyperactivity Disorder (ADHD) (So¨derlund et al., 2007). Current study addresses the issue whether noise can improve response inhibitions in Children with ADHD and its relation to symptoms of ADHD. Children and adolescents were recruited in the study and were assigned into groups according to predominantly inattentive presentation (ADD) or combined presentation of additional hyperactivity-impulsivity criteria (ADHDcom). We adopt the visual stop signal task (VSRT) for inhibition function. Participants performed VSRT under levels of background noise (0dB、35dB、55dB、80dB). The results revealed that 55 to 80 dB background noise selectively improves performance of ADD relative to ADHDcom group. Given that inhibition is the critical problem of ADHDcom, moderate noise is enough to selectively improve that of ADD rather than ADHDcom.

ABSTRACT. We investigated the effects of the inhibition of eye movement toward the cue on letter recognition processing. The letter (O or Q) was presented in the peripheral visual field as a target stimulus. Participants were required to judge which target letter appeared and to press the corresponding (O or Q) button. In the cue condition, before the target appeared, a square with the same letter as the target written on it was presented in the peripheral visual field as a cue stimulus. In the no-cue condition, the target letter was not written in the square. Participants were required to inhibit their eye movement when the square was presented. The position of the square (left or right), the duration of the square presentation (0.1, 0.4, 0.7, 1.0 s), the target letter (O or Q), and the position of the target (left or right) were randomly determined. The results showed that the reaction time was significantly faster with a longer duration in the cue condition. This suggests that participants took advantage of the cue information during the cue period in which they were required to inhibit their eye movement, because they could recognize the cue letter with presented for a longer duration. Moreover, the reaction time for the letter presented for a 0.1 s duration, wherein participants could not recognize the cue letter, was significantly slower in the cue condition than that in the no-cue condition. This result indicated that the presentation of a cue had a negative effect on participant performance. Taken together, these results suggest that participants made efficient use of cue information during the inhibitory period and performed better when they could take advantage of the cue information, while performance would suffer when it was difficult to obtain cue information.

ABSTRACT. There have been many previous studies showing that the direction of other’s eye gaze captures viewer’s attention. However, as most of the studies presented the face on the fixation point, the effect of gaze cue in the periphery is unclear. In this study, two schematic facial stimuli were presented horizontally away from the fixation point at both sides and the effect of position of the face was investigated. The circular facial stimuli which subtended 2.1 deg in diameter were presented 1.0, 1.5, 3.0 or 4.5 deg away from the fixation cross at both sides. There was also a control condition that only one face was presented at the fixation point. After the presentation of frontal gaze for 750 ms, the both eye gaze of two faces or the eye gaze of one face were directed either to the left or right for 300 ms. The circular target which subtended 0.9 deg was presented 6.4 deg away to the left or right from the fixation cross. Participants (N = 10) responded the position of the target (left / right) as soon as possible and also responded the gaze direction (left / right). The results showed that the reaction time for localizing the target was significantly shorter for cued than uncued gaze direction in all position conditions and there was no significant difference in the effect of gaze direction among the position conditions. Additionally, the response time was significantly longer in both cued and uncued gaze direction condition when the face became farther from the fixation point. These results suggest that although the perception of gaze direction itself takes longer time when the face was farther from fixation point, the effect of attentional capture by gaze cue was still significant at least 4.5 deg away from the fixation point.

ABSTRACT. Computational studies of human visual attention suggested that salient locations in a scene (saliency map) can be used for gaze prediction of human observers (e.g., Itti et al., 1998). Previous studies reported that head information together with a saliency map obtained by a visual attention model improves prediction accuracy of gaze locations during observation of static images (e.g., Nakashima et al., 2014). The method was based on the fact that eye and head move cooperatively (Nakashima et al., 2012). In the present study, we investigated whether the method can be used to predict gaze shifts during watching a movie sequence. We estimated the likelihood of eye positions with respect to a head orientation based on the data provided from the previous study in which human observers watched an Ultra High Definition movie (Fang et al., 2015). We used a two-dimensional (2-D) Gaussian distribution to approximate eye positions with respect to a head orientation and the estimated eye distribution was multiplied with saliency map, yielding a gaze-estimation map. We evaluated prediction accuracy as AUC (area under the curve) score of the threshold and hit rate function for three models: (method 1) saliency map, (method 2) estimated eye position (2-D Gaussian), and (method 3) gaze-estimation map. The results showed that the large improvement was found when information of head position was considered (method 2 and 3 > method 1), and that inclusion of saliency slightly improved the performance (method 3 > method 2). These results highlight the important role of head orientation in the deployment of attention.

ABSTRACT. The gist of a scene can be extracted from a brief glimpse. Recently, many eye movement studies have suggested that the initial representation obtained by a glimpse of a scene can facilitate visual search behaviors, e.g., reaction times (RTs) and eye movements, for an object embedded within the scene. In the present study, we investigated the relationships between scene perception and visual search performance. In Experiment 1, we examined whether scene gist could be briefly acquired. In this experiment, one of two scenes (i.e., kitchen room or living room) was presented for 25, 50, or 75 ms. Participants were asked to indicate a category of the scene. The results showed relatively high proportion of correct responses (more than 70%) in all the conditions. The proportion of correct responses increased with increasing exposure duration. In Experiment 2, using the scene images used in Experiment 1, participants performed a visual search task for a target object in the scene. In Experiment 3, participants performed a visual search task with a briefly flashed scene preview that preceded each search display. The exposure duration of the preview was manipulated as 0, 25, or 75 ms. The results showed slower RTs in the absence of the preview (Experiment 2) than in the presence of it (Experiment 3). The RTs were significantly shorter in the 75-ms condition than in the other conditions (Experiment 3). These results support the previous findings that the gist of a scene can facilitate visual search performance. The results also suggest that the minimal time required to process a natural scene sufficiently for subsequent visual search is equal to or less than 75 ms, which is similar to the findings of previous eye movement studies.

ABSTRACT. Visual features such as colour, orientation, luminance, eye-of-origin information are reported to modulate attentional deployment and thus visual search. Among them, the effect of ocular origin of information on visual search performance is less clear: Zhaoping (2008, 2012) found that an ocular singleton improves visual search performance when it was the target, while we (Yeung et al, 2014) found consistent search impairment by ocular singleton/column targets across experiments, except in one condition when a luminance-defined distractor was also presented. In this case, ocular singleton/column had no effect on target discrimination. To further examine how ocular information interacts with luminance information, here we addressed two concerns related to the previous experiment: (1) imperfect binocular fusion, and (2) ocular effect masked by high salience of the luminance feature.

We adopted a search display similar to our previous one: one eye viewed a 9x9 vertical-grey-bar-array in mid-grey against dark-grey homogeneous background. There was a randomly-selected luminance-defined singleton column of light-grey. The other eye viewed another independently selected singleton column. We (1) assisted binocular fusion by including a binocular outer ring. For (2), we adjusted the contrast of the luminance-defined singleton column against background at 4 levels to lower its feature salience.

After we lowered luminance salience, we discovered consistent search impairment by ocular singleton column that was not apparent before. Our result suggests that eye-of-origin information was available to observers during search, and an ocular singleton column hinders target detection, similar to the effect of a collinear singleton column (e.g. Jingling & Tseng, 2013). Our finding highlights the possibility of collinearity and eye-of-origin sharing common mechanisms in influencing attentional allocation in visual search.

ABSTRACT. When localization responses are made in a cueing paradigm, inhibition of return (IOR) falls into two categories depending on whether the oculomotor system is activated or suppressed. When saccades are allowed, output-based IOR is generated regardless of whether they are made to cues or targets and whether stimuli are exogenous (peripheral) or endogenous (central arrows; except when there is no response central cues). When suppressed, input-based IOR arises only when targets are exogenous. It has been suggested that output-based IOR is the ‘true’ form of IOR thought to facilitate foraging and novelty seeking, whereas input-based IOR is related to habituation or adaptation processes. In a series of experiments, we mixed pro- and anti-localization tasks in order to investigate whether cueing effects would follow the same pattern as those observed in the traditional paradigm. In all experiments, an uninformative cue was displayed, followed by a cue-back stimulus that was either red or green. Green cue-backs indicated that pro-localization responses were to be made to the upcoming targets, whereas red called for anti-localization (respond opposite the targets). In the first experiment, subjects did not make any saccades, whereas saccades were made to cues in Experiment 2, and to targets in Experiment 3. We predicted that pro-localization trials would show the same pattern of effects as previously observed with peripheral stimuli (IOR in all three experiments), whereas anti-localization trials would result in similar performance as with central arrows (IOR when activated and no cueing effect when suppressed). However, our results showed that IOR was only observed when saccades were made to the targets (in both conditions), never with manual responses. This suggests a dissociation in the generation of cueing effects depending on response modality when anti-localization responses are required.

ABSTRACT. In recent years visual displaying apparatuses are getting bigger. However, the way to produce visual contents for wider screen seems unchanged. In order to produce visual contents optimal to wider screens, it is important to know the behavior of our visual system toward it. Therefore, we investigated the effect of the size of visual field toward which visuospatial attention is oriented on the brain activity, particularly even-related potentials during visual search. Methods. The experimental task was the visual search. The stimulus consisted of 66 oriented line segments positioned on six concentric circles whose radii ranged from 1.6 to 25.6 degrees. Length of the segments varied from 1.7 to 13.4 degrees in accordance with cortical magnification factor. The orientation-defined target was embedded in distractors with uniform orientation. In every trial, following a cueing stimulus (2 secs duration and 8 secs blank) for area to be attended the task stimuli (0.5 sec duration and 2.5 secs SOA) were presented six times. The radii of the attended areas were 3.9, 10.7, or 32.3 degrees. By modulating target-distractor orientation difference task difficulty was fixed across attended area conditions. EEG signals were recorded from electrodes at Cz, Pz, P3, P4, T3, and T4 referred to an electrode at nasal root. The signals were re-referenced to the average of A1 and A2 and were amplified using a 0.1-30 Hz bandpass filter. Results and Discussions. Event related-potentials for correctly responded frames were calculated for each condition. N2 component (270-320 msec) showed significant area size-channel interactions. Especially parietal channels (Pz and P3) showed the trend that expansion of attended area decreased N2 amplitude. Several past studies suggested relationship between N2 component and cognitive control or attention. Therefore the current result would suggest that attending very wide visual field may deteriorate some kind of higher order visual cognition.

ABSTRACT. The receptive fields of the clear majority of neurons in the primary visual cortex (V1) of the cat contain silent surround regions beyond the classical receptive field (CRF). When stimulated on their own, the silent surround regions do not generate action potentials (spikes); instead, they modulate (usually partially suppress) spike responses to stimuli presented in the CRF. In the present study, we subdivided our sample of single V1 neurons recorded from anesthetized cats into two distinct categories: surround-suppressive (SS) cells and surround-non- suppressive (SN) cells. Consistent with previous reports, we found a negative correlation between the size of the CRF and the optimal spatial frequency (SF) of circular patches of achromatic gratings presented in the cells' receptive fields. Furthermore, we found a positive correlation between the strength of the surround suppression and the optimal spatial frequency of the achromatic gratings presented in the cells' receptive fields. The correlation between the strength of surround suppression and optimal spatial frequency was stronger in neurons with suppressive regions located in the so-called 'end' zones. The functional implications of these relations are discussed.

ABSTRACT. Inhibition of return (IOR) refers to a bias against previous attended locations. This attentional mechanism is frequently explored with either cueing tasks or oculomotor search tasks. With cueing tasks, previous studies have shown that dyslexic children have deficits in IOR (e.g., Facoetti et al., 2003). In the present study, we further assessed IOR in dyslexic and normal developing children with a cueing task that required covert shift of attention and an oculomotor search task that required overt shift of attention. Reading tests were first carried out to select a group of dyslexic children (6 females and 9 males, mean age: 9.3 years old) and a group of age-matched controls (4 females and 13 males, mean age: 9.7 years old). These children then completed both a cueing and an oculomotor search tasks. In the cueing task, children responded to peripheral onset targets that were preceded by uninformative cues. In the oculomotor search task, children searched for a camouflaged target from 30 highly cluttered images. An Eyelink 1000 eye tracker was used to monitor children's eye movements in both tasks. In the cueing task, the results revealed a reliable IOR effect for age-matched controls (15 ms) but not for dyslexic children (-7 ms). In the oculomotor search task, IOR is manifested by prolonged latency for saccades returning to the vicinity of previously fixations. This effect was observed for age-matched controls (11 ms) but not for dyslexic children (3 ms). Taken together, the present findings suggest that dyslexic children may have a general IOR deficit in both covert and overt attention tasks

ABSTRACT. Inhibition of return (IOR) is a bias against recently attended locations. Previous studies have shown that IOR is not restricted to previously attended locations, but rather has a spatial gradient. However, little is known about the temporal characteristics of the IOR gradient. The present study was set out to fill up this gap with a modified cueing paradigm. Because the IOR gradient is fairly large, to measure IOR effects in cueing tasks, it is necessary to use baseline locations that are far enough from the cued location. In the present study, a set of three monitors were used, with one in the center for presenting the fixation stimulus and two others on each side for presenting the cues and targets. The visual target was a bright disk presented on the horizontal meridian, 40° left or right to the fixation. The cue was bright box that could appear on the same monitor as the target or on the monitor opposite to the target. There were 21 possible cue locations on each monitor and the distance between the cue and the target, if they appeared on the same monitor, could be 0-15°, in steps of 3°. To reveal the temporal characteristics of the IOR gradient, the cue-target onset asynchrony (CTOA) was varied across 100 ms, 400 ms and 1200 ms (intermixed within block of trials). The results show that a) the magnitude of IOR reaches its maximum at a CTOA of about 400 ms and then decreases, and b) the size of the IOR gradient increases as the magnitude of IOR decreases. These observations seem to suggest that once IOR is fully developed the total amount inhibition will remain constant for a while, the weakening of IOR is mainly the result of its spreading in space.

ABSTRACT. People tend to automatically shift their attention to the same object that other people are looking at or an arrow is pointing at. Though plenty of researches have been conducted, similar shifts of attention were usually reported for gaze cueing and arrow cueing. Different from previous studies that used between-block designs, the present study compared the attention orienting between gaze and arrow cues in a within-block design. Top-down processes, such as direct comparison of cue types, are available in a within-block design. In experiment 1, previous undistinguishable results were replicated under a between-block design. In experiment 2, a within-block design was used and the central cues (gaze or arrow) changed from trial to trial. In experiment 3, the central cues were chosen randomly within one block and two cue-target SOAs were included. As a result, gaze-induced cueing effects are stronger than arrow-induced cueing effects in both experiment 2 and experiment 3. Furthermore, the stronger cueing effects for gaze cues than arrow cues are only observed at relatively long SOAs. The present findings suggest that the attentional mechanisms induced by gaze and arrow cues differ in higher cognitive processes, and eye-gaze triggers attention orienting more effectively than arrow cues.

Supported by NSFC(31300938,11403009,and 61462052) and Scientific Research Fund of Yunnan Provincial Department of Education(2013Z134).

ABSTRACT. Contextual effect refers to the extent to which people are influenced by contexts. After reviewing sex differences for various domains, this study proposed a hypothesis that women are more sensitive to contexts than men, which was tested by using a figure-ground congruency paradigm and a contextual cueing paradigm. The participants in this study were university students and third- and fourth-grade elementary school students. This study found that female university students achieved a higher correct rate when the target and background were congruent than when the two were incongruent. In addition, compared with invalid contextual cues, female university participants obtained a higher correct rate when the contextual cues were valid. In comparison, male university students were influenced by contexts to a relatively lesser extent. Therefore, the results shown by university students supported the hypothesis. The congruency effect and contextual cueing effect showed sex differences in the reaction time performance of third- and fourth-grade elementary school students. However, the direction of sex differences exhibited by students in various grades varied according to the types of targets. This result should be compared with that obtained from adults for further analysis.

ABSTRACT. It has been argued that seeing shapes entails seeing time. A bitten cookie, for example, is seen as a whole cookie that was subsequently bitten. Similarly, a twisted towel is represented as an untwisted towel that was subsequently twisted. Do such inferences arise during visual processing itself? To explore this, we tested whether the perception of history in static shapes could trigger the perception of motion. Observers were told that they would see short movies of a shape (e.g. a square) changing from its complete form into a truncated form, with a "piece" of it missing, and that this change could occur in two ways: (1) all at once, in a single flash; or (2) gradually, with the missing piece quickly "growing" into the shape (as when you poke your finger into a lump of clay). We manipulated the contours of the missing piece to look either as if another shape had "intruded" on the original shape, or as if there had been no such past gradual transformation. When the shape suggested an intrusion, sudden changes were much more likely to be (mis-)perceived as being gradual, in a type of transformational apparent motion. This effect was exceptionally robust, and occurred reliably in every observer. Additional experiments demonstrated that this effect: (1) was due to perception and not response biases (since the effect did not persist with real gradual changes); (2) was not due to differences in lower-level visual features (since it replicated while controlling for geometric features such as area and opening width); and (3) was due to perceived causal history per se (since the effect disappeared when the very same shapes were presented so as to suggest occlusion rather than intrusion). In sum, the perception of causal history in static shapes is powerful enough to induce illusory motion percepts.

ABSTRACT. Kitazaki and Sato(2003) have reported that non-attended rather than attended motion dominates vection when two opposing motion exited in the stimulus. Recently, Seno et al. (2009) reported that with a homogenous full-field(single) motion, vection is weakened when attention was deprived by simultaneous execution RSVP/MOT tasks. These two results apparently contradict to each other in respect to the role of attention. To resolve the contradiction, in the present study, we compared results for one and two opposing motions stimuli with and without attentional load. First, we examined the effect of attntional task in Kitazaki and Sato(2003) paradigm. The stimulus was a vertical grating pattern comprised of red- and green-dots (121Vx61H deg, stripe-width 7.5 deg). Red- and green-dots moved in opposite directions(upward/downward). Observers paid attention to one of the motions, i.e. colors, and judged the direction of perceived vection by pressing a button. The latency, duration, and subjective strength (magnitude) were measured. In the dual-task condition, they simultaneously executed an RSVP task while observing vection stimuls. The RSVP stimulus was presented at the center of stimulus. It was found that the vection was dominantly induced by unattended motion in both single- and dual-task conditions. That is, the deprivation of attention by RSVP does not affect at all to the vection discrimination. However, the strength of vection was somewhat weaker in dual-task condition. Next, we examined the vection strength in single- and dual-task condition by using a homogeneous stimulus. For this, vection was significantly reduced by RSVP execution. These results suggest that attention affect vection in two separate levels. First, input motions are sorted as attended (figure) and unattended (ground) motions, and the unattended (ground) motion dominates the input for vection processing stage, where attentional loads has negative effect on probably decision making and produces reduction of the vection strength.

ABSTRACT. Patients who suffer from striate cortex (V1) damages can retain unconscious visual functions within the affected parts of their visual fields (“blindsight”). Blindsight is more common in patients who sustained V1 lesions in childhood. However, physiological and anatomical studies of residual vision in primate models have concentrated on animals with lesions introduced in adulthood. The consequences of early-life V1 lesions remain largely unknown. To examine if younger brains have greater capacity for reorganization, we characterized the response properties of neurons in the Middle Temporal area (MT, a major recipient of V1 efferents), and the lateral geniculate nucleus (LGN, a major source of V1 afferents) in marmoset monkeys with partial unilateral V1 lesions placed within 6 postnatal weeks. Extracellular recordings were performed after one year of recovery. Unlike in adult-lesioned marmosets, where a small population of MT neurons can be found to respond weakly to stimuli in the affected visual field, we found that virtually all neurons in the region of MT that were deprived of V1 inputs showed robust responses to visual stimulation. In addition, the normal visutopic organization of MT was maintained. However, direction selectivity, a key physiological characteristic of MT which is known to be preserved in many cells following adult V1 lesions, was absent - suggesting that the activity of V1 in early life is critical for the development of direction selectivity in MT.

In the LGN, although the majority of the neurons died following V1 lesion, neurons could be found scattered throughout the degenerated region. LGN neurons in the degenerated zone responded robustly to visual stimuli, with normal-sized receptive fields following the canonical LGN visuotopy and layer-specific eye-dominance. The results suggest that LGN could be a source of MT activation in the absence of V1.

ABSTRACT. The middle temporal area (MT) is considered the center for motion processing in the primate brain. However, the neural mechanism for encoding the speed of a moving stimulus at the level of MT remains largely unknown. Single-unit studies which use drifting gratings to characterize speed tuning properties have suggested that the vast majority of MT neurons are tuned to the temporal frequency the stimulus, indicating that they cannot encode speed unambiguously. Here, we demonstrate with electrophysiological recordings in the brains of anaesthetized marmoset monkeys that earlier experiments may have greatly underestimated the capacity of MT neurons to encode speed. The extents to which MT neurons encoded the speeds of sinusoidal patterns despite variations in their spatiotemporal frequency contents were measured by independent manipulation of their spatial frequencies and speeds. In addition to the traditional drifting grating stimulus, which modulates the spatial phase of the carrier wave while its envelope remains stationary, we also characterized the same set of neurons with a novel stimulus design, consisting of Gabor patches that traversed across the visual field in trajectories. Replicating earlier studies, few (1.8%) among 110 MT neurons that we sampled were speed-tuned when tested with drifting gratings. However, a significantly larger proportion (50.0%) of the sample neurons showed speed-tuning when tested with the moving stimuli, despite the fact that the same combinations of spatial and temporal frequency were presented at the centre of the receptive fields in both conditions. This result suggests that motion of a discrete object across the visual field, as found in most natural situations, could elicit a distinct set of computations in area MT, leading to more robust encoding of speed.

ABSTRACT. A geometric design comprising a surrounding region of oblong checks and an inner circular region of the same checks with a different orientation (Ōuchi, 1977), produces a striking illusion of relative motion whereby the inset is seen to separate from the background and produce slight ‘jiggling’ or sliding motions in random directions in a desultory way. A recent review by Spillmann (2013) highlighted that although the illusion captivated all over ten years ago, little progress was and has been made on an explanation of the effect. We conducted a series of experiments using naïve observers exploring spatial (spatial frequency and angles) and temporal (velocity of movement and rate of change of direction) parameters affecting the strength of the illusion to circumscribe an explanation. To more easily examine these parameters, a simple stimulus comprising two sinewave gratings (one surrounding the other) was used. This does produce a strong illusion of relative motion similar to the Ōuchi pattern. The entire stimulus underwent small, rapid oscillation for 0.25 or 1.0 sec behind a fixation point. All parameters contributed to determining the strength of the illusion. The illusion was most potent for high oscillation frequencies (6 Hz), velocities around 2 deg/sec or less, obtuse angles between local motion vectors (120 deg or more) that were oriented horizontally and vertically, and spatial frequencies that were dependent upon the size of the stimulus (from 2 to 7 cpd). An explanation based on the resolution of the aperture problem is proposed which assumes that at least 200 msec is required for the true direction of motion of terminated lines to be resolved as does small angles between a contour and the direction of motion especially when the motion signal is weak.

ABSTRACT. It has been debated whether classical apparent motion is detectable by either a first-order motion mechanism (Cavanagh & Mather, 1989) or by an attention-based mechanism (Horowitz & Treisman, 1994). We reported that, when classical apparent motion stimuli consisting of two discs were successively presented within or across hemifields (right/left or upper/lower), the motion perception rate declined at shorter ISIs in cross-hemifield conditions relative to within-hemifield conditions (Sato, Kanaya, & Fujita, VSS2013). These results suggest classical apparent motion is partially mediated by a first-order-like motion mechanism or by another lower-motion mechanism. To further clarify this point, in this study, we examined the interocular display effect of classical apparent motion stimuli (Shipley, Kenney, & King, 1945). Two white discs (1.02 deg diameter) were successively presented on the gray background either by binocular display (presentation to two eyes), monocular display (presentation to one eye—left or right), or interocular display (separate presentation to different eyes), using a head-mounted display device (HMZ-T2, SONY). ISI was varied in seven steps between 0 and 533.3 msec. The observer’s task was to judge whether motion was perceived or not. It was found that apparent motion perception had a tendency similar to those of Sato et al. (2013), and was much the same in three display conditions. These results indicate that a lower-motion mechanism other than the first-order motion mechanism, e.g., long-range process (Braddick, 1974, 1980), mediates classical apparent motion.

ABSTRACT. We ordinarily think of vision as something that occurs inside the head.  But it also provides a sensory coupling to the outside world, creating a larger agent-environment system.  In this talk, I will develop the behavioral dynamics view that stable, adaptive behavior emerges from the dynamics of this coupled system.   

At the individual level, consider visually-controlled locomotion through a cluttered environment.  Based on experiments in virtual reality, we have developed a pedestrian model of this locomotor behavior, including steering, obstacle avoidance, and pedestrian interactions.  Locomotor trajectories emerge on-line from the agent-environment interaction, without appeal to an internal world model or explicit path planning.

At the collective level, crowd behavior is believed to emerge from local interactions between individuals.  Indeed, simulations of our pedestrian model reproduce patterns of crowd data from key scenarios like Grand Central Station, Swarm, and Counterflow.  Individual trajectories and crowd dynamics can be captured with just a few basic behaviors. 

At both levels, behavior thus emerges from local interactions, without invoking internal models or plans, consistent with principles of self-organization.

ABSTRACT. In the classic contextual cueing paradigm, participants performed a visual search task in which they looked for a target letter T among a number of distractor letter Ls. Half of the scenes were repeated over blocks while the other half were novel. Over time, smaller reaction times were found for repeated scenes than for novel scenes (contextual cueing effect, Chun and Jiang, 1998). Recent research has shown that, following an initial learning phase in which a reliable contextual cueing effect had been established, when the target locations were exchanged between two different repeated scenes, the contextual cueing effect was largely maintained (Zellin et al., 2013). Instead of the maintenance of contextual cueing effect, the current study examined whether contextual cueing effect could be generated without a fixed one-to-one relation between target location and context. We used a modified contextual cueing paradigm in which target locations were exchanged between repeated scenes across blocks during the learning phase. For the repeated scenes, in each block there were total of 8 target locations and 8 contexts. For a given target location (or context), the number of possible pairing between target and context was varied among four levels: 1 (control), 2, 4 and 8. When each of two target locations were paired with each of two different contexts (over two consecutive blocks), comparable contextual cueing effect was found as that of the control. Moreover, there was an inverse relation between number of possible pairings and magnitude of contextual cueing effect. The results suggest that one-to-one relation between target and context is not the necessary condition for contextual cueing effect, but contextual cueing effect would be greater for smaller number of possible pairing between target location and context.

ABSTRACT. In a visual search task, participants search faster for repeated scenes than for novel ones (contextual cueing effect, Chun and Jiang 1998). Our earlier eye tracking study demonstrated that this RT benefit could be attributed to both attentional guidance and response selection (Zhao, et al, 2012). In a recent study, we used a modified contextual cueing paradigm in which two target locations were exchanged between repeated scenes across blocks. Comparable contextual cueing effect was found as that of the control. In the current study, using the same paradigm, we varied the set size of the distractor items to identify which phase of the task that lead to the RT benefit. In the experimental condition, over two consecutive blocks, a given target was paired with two different repeated contexts and a given repeated context was paired with two different targets. The control condition used the classical contextual cueing effect paradigm with the same one-to-one relation between target and context in each block. In each of the two conditions, participants searched a target in the distractor array of two different set sizes: 12 and 16. We found contextual cueing effect in both conditions. For the function of RT vs set size, the slope for the repeated scenes was lower than that of the novel ones only for the control condition, but not for the experimental condition. The intercept for the repeated configurations was much smaller than that of the novel ones only for experimental condition, but not for the control condition. These results suggested that attentional guidance might contribute to the contextual cueing effect in the traditional contextual cueing paradigm, but response selection may play a major role in contextual cueing effect where multiple relations exist between target location and context.

ABSTRACT. In a visual search task, if target appears more frequently in one region of the scene, over time, participants will implicitly learn the target location probability. This has been demonstrated through faster search when the target appears in the high probability region compared to regions less frequently occupied, and such faster response remained even when, later on, target location probability was changed to be equal throughout the entire screen. Previous research, using letter search task, has demonstrated that the attended locations are viewer centered and are not updated with viewer movement (Jiang & Swallow, 2012). In this study, we reexamined the spatial frame of reference in this type of learning using computer rendered illustrations of 3D scenes. The scene consisted of a display of an array of chairs randomly positioned on the ground but with coherent orientation. Participants searched for and identified a target positioned on a chair. Participants learned the layout from one or two views and, before testing, moved to a different view. During training, across multiple trials the target was more often found in one, rich quadrant than in any one of the sparse quadrants. In the testing phase, the target was randomly placed appearing in each quadrant with the same probability. Results demonstrated probability cuing in the training phase indicated by decreased reaction time in the rich quadrant compared to the sparse quadrants. In the testing phase, most participants showed lowest reaction time in the quadrant that maintained the same spatial relationship with the viewer as the previously rich quadrant. However, small proportion of the participants also showed comparably low reaction time in the original rich quadrant. These data suggest while for most participants the attended locations are viewer centered without update with viewer movement, small percentage of participants could possibly update with their movements.

ABSTRACT. Introduction: The frontal eye field (FEF) and the dorsolateral prefrontal cortex (dl-PFC) are two prefrontal regions that share direct reciprocal anatomical connections, and are thought to be involved in the planning and execution of saccadic eye movements and the control of visual selective attention. Despite the clear integration of these two regions in visuomotor processing, no studies to date have analyzed the activity of both FEF and dl-PFC simultaneously to understand the causal relationships between them. In this study, we analyzed the causal relationships in local field potentials (LFPs) between FEF and dl-PFC when an animal performed a memory guided saccade task.

Methods: Neuronal signals were recorded in an awake-behaving monkey (Macaca fusicularis) using chronically implanted electrodes while the animal was performing a memory guided saccade task. 32 electrodes were implanted in FEF and 48 electrodes in dl-PFC. The raw signals were low-pass filtered with a 150 Hz cut-off frequency to obtain the local field potentials. In this study, Granger Causality was used to analyze the interactions between dl-PFC and FEF. A sliding window of 250 ms, with a step size of 50 ms, was used to compute the magnitude and direction of Granger Causality.

Results: The majority of dl-PFC channels, and some of the FEF channels, have selectivity in the beta and low gamma frequency range (20 Hz to 50 Hz) during the delay period. The directed coherence values computed from the dl-PFC to FEF were significantly higher when compared to directed coherence values from FEF to dl-PFC. The directed coherence for LFPs with overlapping receptive field locations were significantly higher than locations that were outside the receptive fields of both channels. The results showed that dl-PFC appeared to have a stronger influence on FEF and the latency in the causality between the two areas was about 12 ms.

ABSTRACT. Background: The frontal eye fields (FEF) and the dorsolateral prefrontal cortex (dl-PFC) are two prefrontal regions that share direct reciprocal anatomical connections, and are thought to be involved in target selection, planning and execution of saccades. In addition, both FEF and dl-PFC cells have been shown to be active during the maintenance of working memory. In this study, we identified a novel response property in a sub-population of cells that showed selectivity to a distractor location only when the previously presented target was at another location. Methods: The activity of single neurons was recorded in an awake-behaving monkey (Macaca fusicularis) using chronically implanted electrodes in FEF and dl-PFC while the animal performed a visual memory guided saccade task. In our task, the presentation of the target was followed by an initial delay period before a visual distractor that the monkey was expected to ignore was presented. This was followed by a second delay period, before a response cue indicated the start of the response period. Results: We found a previously unidentified response property in a sub-population of prefrontal neurons. These cells responded selectively to distractor location only when the previously presented target was at another location. This activity was seen during the distractor presentation period, as well as the delay period that followed it. FEF showed the maximum number of cells with this activity during the distractor presentation period (14.7%). On the other hand, 29% of dl-PFC cells showed this activity during the delay period that followed the distractor presentation. In 60% of these cells, the distractor was presented at a location adjacent to the target. Finally, the latency of this response during the distractor presentation period was higher than the latency of a visual response, suggesting that the response was not merely of visual origin.

ABSTRACT. Visuomotor coordination involves two internal models: the forward model and the inverse model. Major theories explain that two types of error lead visuomotor learning for internal models. Predictive error (difference between predicted effector positions and actual effector positions) updates the forward model while feedback error (visually perceived discrepancies between actual effector positions and target positions) updates the inverse model. The previous studies indicated that the predictive error solely did not enhance visuomotor adaptation and it required feedback error as well. It was not clear, however, that the predictive error component of the combined error actually contributed to the acquisition of a new tool manipulation. This study examined whether the forward internal model acquires in the combined error condition. The experiment, therefore, set the two conditional phases, (1) the visible cursor phase (VC), and (2) the invisible cursor phase (IC). In the ten trials of a training session with VC, participants were asked to track a visible target to acquire new internal models. In the test phase, they continued tracking with a visible cursor first five seconds (first epoch), then without a cursor (second epoch) for ten seconds. Those who had acquired forward internal model completely during the VC would show the same performances under the IC. Individuals without such complete forward internal model would show worse tracking performances in the IC than in the VC. The result showed that the performance in IC was lower than in VC, and, the correlation of the performance between in IC and in VC was weaker in the second epoch (r=0.28) than in the first epoch (r=0.52). Therefore, the predictive error component of the combined error contributes differently to the acquisition of internal models. Performances of tool-usage would not good predictor of tool’s forward internal model acquired by manipulator.

ABSTRACT. When an object go through a cast shadow, the lightness of the object becomes darker. We could judge easily the lightness changes caused by illumination changes, but not by changes of surface’s reflectance, because we have the assumption that the lightness in a shadow would be darker than out of that. In the current study, we investigated whether infants use such assumption to perceive the lightness of the object in shadows. We presented infants two 3DCG movies: in natural condition, a chick go across cast shadows of two panels and the luminance of the chick became darker when it passed the cast shadow naturally; in unnatural condition, the luminance change of chick’s surface was reversed, lighter in the shadows and darker outside of shadows. If the infants use the “shadow assumption” to perceive the lightness in shadows, infants would detect the unnatural lightness change in shadow. We used the familiarization method same as Csibra (2001). First, infants were familiarized with a chick moves right to left and it was not through the cast shadows. After this familiarization, we presented the natural and unnatural condition one trial respectively. Csibra, (2001) showed that infants would prefer to something unnatural. In results of our study, only 7- to 8-month-olds looked longer the unnatural condition than the unnatural condition same as Csibra’s study. The results revealed that the 7- to 8-month-old infants could detect the unnatural changing of the luminance of chick’s surface, but the 5- to 6-month-old infants could not. These results suggested that 7- to 8-month-old infants would perceive object’s lightness by using the assumption that the lightness in a shadow would be darker than outsides.

ABSTRACT. In a recent work (Gao et al., Color Constancy Using Double-Opponency, IEEE Trans on PAMI, 2015), we simulated the properties of a kind of double-opponent (DO) cells with concentric receptive field (RF) in the primary visual cortex (V1), and found that when the RF center and its surround receive unbalanced cone inputs, the outputs of DO cells could be used to accurately estimate the light source color. Computationally, unbalanced center-surround structure can emphasize the edge information of scenes. In this work, under the help of eye-tracker (Eyelink-2000, SR Research Ltd), we further investigated the role of different local information of scenes in human color constancy. For each of the 16 subjects, each of the 40 Mondrian images was displayed randomly on the screen; each image was displayed 10 times, illuminated (multiplied) with a light source of a random color. The subjects were asked to remove the illuminant color from the color-biased scenes by press the up- or down-arrow keys. In addition, free-viewing condition was introduced to act as control task. The results show that (1) more fixations are distributed around the local region edges, as suggested by our computational model mentioned above; (2) more fixations are located in the achromatic local regions (e.g., white or grey regions). In short, achromatic and edge information of scenes seem to contribute more for human color constancy. [Supported by grants of 973 Project #2013CB329401 and NSFC #61375115, #91420105]

ABSTRACT. The stimulus–response associations acquired by learning enable us to respond quickly to visual stimuli. Ishizaki et al. (2015) studied how integrated visual features are represented and associated with responses by using multi-attribute stimulus–response mapping tasks. They used shape, color, and texture as the stimulus attributes, and the results supported the paired-attribute model in which bound feature pairs, rather than unified object representations, are associated with responses.

As different visual attributes are thought to be bound together by focusing attention on the location of the object, the location attribute is considered to play a special role in the attentional binding process. Thus, we examined whether location is different from other attributes in terms of how it affects the attribute integration process of stimulus–response mapping.

We conducted experiments on the following task: subjects learned the mapping from eight stimulus items (combinations of two different colors, shapes, and locations) to four response keys. Two items of the same shape and color but at different locations (the shape–color set) corresponded to key A. Similarly, the shape–location set and color–location set corresponded to keys B and C, respectively. Two remaining items of different shapes, colors, and locations from each other corresponded to key D.

As a result, learning and performing the task was more difficult when three attributes determined the correct response (key D) than when two attributes did (keys A, B, and C), which accords with the paired-attribute model. Moreover, the shape–location and color–location sets tended to be more difficult to learn than the color-shape set, though location was easier to detect than shape or color. The results suggest that integrated representations of shape–location and color–location, as well as shape–color (but not shape-color-location) representations, are formed to be associated with responses.

ABSTRACT. In the human visual system, different attributes of an object, such as shape, color, and texture, are first separately processed and then integrated to produce a specific response. Previous research has suggested that attentional binding of these attributes contributes to this process but unified object representations containing all attributes are not used. To confirm this hypothesis, we examined the effects of spatially separate presentations of different attributes, which inhibit the binding between them, on a multi-attribute stimulus–response mapping task. In this task, subjects were required to associate eight stimuli (combinations of two shapes, two colors, and two textures) with four response keys and to press the correct key as quickly as possible when one of the eight stimuli was presented. The results showed that the response selection time was longer when three attributes were individually presented in different windows than when they appeared in the same window, whereas no significant difference was found when they were presented in pairs, i.e., when shape and color, color and texture, and texture and shape were presented in three windows. In addition, learning and performing the task were more difficult when the response was determined by three attributes than when it was determined by two attributes, suggesting that response selection was not based on the unified representation of three attributes. These results support a paired-attribute model in which bound feature pairs, rather than unified object representations, are associated with responses and strengthen the “no-triplet” hypothesis that attention does not directly bind more than two attributes.

ABSTRACT. The present study aimed to understand the effect of inter-items similarity on visual short-term memory (VSTM) performance. A change detection task was used for testing memory performance, and the number of memory items was manipulated for four (in Experiment 1) or six (in Experiment 2) items. In both experiments, memory items were paired and sequentially displayed according to the level of similarity between the items in each pair. The inter-item similarity between the paired items was made low by selecting their colors from two completely different color categories, whereas was made high by selecting two distinctive colors from a single color category. About a second after the memory item, a probe display designed for a whole or partial-probe procedure followed. Participants reported if the probe display has an item differing its color from the corresponding one in the memory items or not. The results showed that the accuracy of change detection decreased if the inter-item similarity was high compared to when it was low, exclusively in the partial-probe condition. These results suggest that VSTM performance can be impaired if the similarity across the items in VSTM is high, especially when the assessed similarity becomes explicitly available during the retrieval of memory items for recognition.

ABSTRACT. Iconic memory (IM) is a form of storage system with a greater capacity, capturing incoming sensory information to gate them into visual short-term memory (VSTM) limited in its storage capacity. Since Sperling’s experimental demonstration of the transition from IM into VSTM, VSTM’s storage capacity and precision were well understood regarding the time course of representational process. However, there have been few studies of IM for the same topic of interest despite its important role as the first storage system to provide a core set of information for VSTM. Accordingly, we attempted to characterize the temporal characteristics of IM regarding its capacity and precision, and contrasted them against those of VSTM. In our Experiment, two or eight colored boxes were displayed for 100ms, followed by a retro-cue designating which of the boxes must be recalled for its color. Participants selected the color of a cued item on a color-wheel, and the recall performance was assessed in accuracy (Pm) and precision (s.d.) of the selection according to Mixed model analysis (Zhang & Luck, 2009). Critically, the retro-cue was displayed 30, 100, 300 or 600ms after the offset of the boxes. The estimated Pm was near-ceiling in the setsize 2 condition regardless of the cue intervals whereas gradually decreased as the cue interval increased in the setsize 8 condition. The s.d. was on average smaller in the setsize 2 condition than the 8 condition, but in the setsize 2 condition specifically, it increased until the interval of 100ms and stayed constant since then. The results suggest that IM is relatively precise with much bigger capacity compared to VSTM but becomes unavailable almost instantly upon the offset of memory items. These support for the classic view of IM that the limited temporal availability of iconic representation triggers the transition of IM into VSTM.

ABSTRACT. 1. Background/Objective GlassesOff is a commercial off-the-shelf (COTS) perceptual learning software application designed to improve near vision in presbyopes by training the brain to sharpen blur images taken by the eye.

The objective of this study was to evaluate the efficacy of perceptual learning using the GlassesOff software application administered through an iPod to improve the near vision of subjects with presbyopia.

2. Study design/methods A total of 27 treated participants with a mean age of 48.6 ± 3.5 years old were included in the study. 12 were in Group A while 15 were in Group B. Group B which received the treatment only after Group A had completed, served as the controls in this study. Both groups went through repeated assessments to compare their performance before and after treatment. Group A also underwent similar assessment at 3 months post-treatment to assess the retention of treatment effects. 3. Results All treated eyes had improvement in unaided near visual acuity (VA) and contrast sensitivity (CS). In Group A, the average improvement in unaided near VA and reading acuity was 0.15 logMAR (1.5 lines) and 0.8 logMAR (0.8 line) respectively – the mean unaided near VA was 0.21 logMAR (VA 6/9.5) and the mean unaided reading acuity was 0.15 logMAR (VA 6/8.5) post-treatment. The training also improved the unaided near CS, which reached its best-corrected level (1.75 log CS) after treatment. Follow-up data at 3 months post-treatment showed that the gains were reduced by 37.7% for unaided near VA and 18.9% for unaided near CS. 4. Conclusion This study of GlassesOff treatment in participants with presbyopia supports the hypothesis of perceptual learning in its ability to improve uncorrected near visual acuity and contrast sensitivity. Selective implementation of the treatment program to motivated individuals would yield the maximum benefit from this study.

ABSTRACT. Working memory is generally thought to be the ability to retain information in an accessible state, and is often associated with selective attention. Numerous studies have demonstrated that attention select relevant information into working memory. Nevertheless, whether the contents of visual working memory (VWM) also shift visual attention covertly is not yet well understood. Here, we examined the influence of VWM on the metrics of reaction times (RTs) and accuracies (ACC) in two dual-task paradigms. In Experiment 1, participants maintained two or four colored circles as they executed a probe to the orientation of letter ‘v’ appearing on either side of the screen. Before the emergence of ‘v’, a pair of stimuli was presented simultaneously on either side of the screen. One match the sample held in memory, and the other was novel. The results showed that RTs were faster to probes at the location of memory match than no-match locations. In Experiment 2, participants maintained a color in memory as they executed a visual search to the target ‘T’ among distractors. Both the target and the distractors were marked with colored squares. The target was either marked with the memory color (match) or not (no-match). The results showed a more accurate and accelerated response in the match condition than that in the no-match condition. These findings imply that the contents of VWM also modulate the shifts of selective attention.

ABSTRACT. The ability of focusing on task-relevant information while suppressing distraction is critical for human cognition and behavior due to the limited capacity of working memory (WM). We used a delayed-match-to-sample (DMS) task and placed an emotional face distractor in the working memory maintenance interval. Event-related potentials (ERPs) were used to investigate the interaction between emotional face distractors (positive, negative, and neutral faces) and visual WM load (low and high) during both early and late phases of the maintenance interval. Behavioral results showed high WM load decreased accuracy and delayed reaction times. Negative face distractors elicited higher frontal N1 and occipital P1 activity, reflecting rapid attentional alerting and early perceptual processing of negative faces. But the increasing WM load reduces the frontal N1 amplitudes, implying that the higher WM load exhausts limited attention resources and insufficient attention toward distractors. Moreover, an inverse correlation between N1 activation difference (high-load minus low-load) and RTs difference (high-load minus low-load) was found at F3 when viewing negative face distractors, which suggests that frontal areas are more sensitive to resolve the negative distraction. Emotion effect was not found in the late frontal LPP and occipital SNW component. In general, our findings demonstrate that the increasing WM load reduced the emotional distraction by decreasing the perception of distractors in the early phase, rather than by increasing the inhibition of distractors in the late phase.

ABSTRACT. Numeric competence is a crucial skill that undergoes a rapid age-related deterioration; additionally, a sizeable portion of the population exhibits numeric processing impairments. This problem is particularly severe when it involves basic calculations, such as subtractions and additions. There is evidence that parietal cortex is involved in numeric processing, but it is still unknown whether other brain areas are involved, and only a few studies tried to directly assess this by using transcranial random noise stimulation, or tRNS. This type of non-invasive micro-electric stimulation enhances activity in the underlying cortex by increasing neuronal spontaneous activity. We tested the performance in a subtraction verification task across three groups of participants that were stimulated by tRNS over frontal or parietal lobes, plus a placebo condition. Performance was tested both immediately and after seven days. We found that parietal and frontal stimulation significantly improved the performance after seven days, while the placebo condition was always ineffective. These results suggest that, aside the parietal cortex, also the frontal cortex is involved in basic numeric competence, thus opening new avenues for restorative applications of neural stimulation.

ABSTRACT. We recently showed that adapting to dynamic bubbled faces generated significant facial expression aftereffect, but adapting to static bubbled faces did not (Xu et al., VSS 2013, Luo & Xu, APCV 2013). It thus renders the question whether attention plays a role in the above difference. Microsaccade, a kind of unconscious eye movement typically occurs during prolonged visual fixation reflecting covert attention (Engbert & Kliegl, 2003), has been suggested to indicate attention shifts. Five subjects were tested in our study. The task was to judge the emotions (happy vs. sad) of 580 faces from three categories: static full faces, static bubbled faces, and dynamic bubbled faces. We recorded their judgment accuracy of these faces and reaction time. Simultaneously we recorded their eye movement by eye tracker (EyeLink II), and analyzed their microsaccades patterns in particular. We found that although there was no significant difference between judgment accuracy in the three face categories, the subjects spent more time on judging bubbled faces, and more frequent microsaccades toward bubbled faces, with larger amplitude of microsaccades to static bubbled faces. It thus suggests attention shifts in bubbled faces are different than in full faces, and the attention difference in dynamic and static bubbled faces. Our findings may shed light on the mechanism of attention in incomplete face perception.

ABSTRACT. Concerning visual processing, individuals are better at recognising faces of their own- race as opposed to those of other-races (the ‘other-race effect’; ORE). However, concerning verbal processing cross-race differences may not be so apparent: describing a face briefly at post-encoding can enhance recognition, irrespective of the race of the face (e.g., Nakabayashi, Lloyd-Jones, Butcher, & Liu, 2012). We examined whether the facilitative effect of verbalization on the ORE could be due to featural or configural processing which are key cognitive processes distinguishing own- and other-race recognition. Participants were presented with either scrambled (whereby configural information was disrupted) or intact faces during both stimulus encoding and retrieval in an old-new recognition task under control and verbalization conditions. The critical difference between the two conditions was the post-encoding task engaged during encoding. The participants engaged in a filler task in the control condition whereas in the verbalization condition they described each face after it disappeared from the screen. Eye movements were recorded throughout the experiment to examine the effects of verbalization and race of stimulus face on visual attention. The main findings highlighted the importance of configural information during encoding for the benefit of verbalization to arise. In contrast, the size of ORE depended on processing of configural information both during encoding and retrieval. Encoding of intact configural information of a face resulted in stronger ORE, especially when such information was available also during retrieval. Eye movement patterns differed markedly depending on verbalization and race of stimulus face. Our key findings are consistent with literature reporting the role of configural processing in distinguishing own- and other-race face recognition. More importantly, we have shown that configural processing also plays a key role in provoking the beneficial effect of verbalization.

ABSTRACT. Can facial attractiveness be affected by how the external features of the face are masked? We examined this question by manipulating the shape of the mask. In one condition, the hair, the ears, and the neck were removed according to their outlines. In another, these external features were cropped via a predefined oval shape. Participants rated attractiveness of the cropped faces. Results showed that faces cropped by the oval shape were rated more attractive than the same faces cropped by their external feature outlines. To investigate whether this oval-shape advantage was due to the shape’s regular and symmetrical appearance or the fact that it cropped out a larger area of the external features, we added another condition in a second experiment, where the same shape of irregular mask was used to crop out a larger area of the external features. The other two conditions remained identical to the first experiment. The results showed that faces cropped by an oval shape was rated most attractive, followed by those cropped by a small irregular shape, which were rated more attractive than those cropped by a large irregular shape. These results suggest that faces tend to look more attractive when a large area of external features and their irregular shape are masked or hidden. Our third experiment further showed that the oval-shape effect was enhanced when facial attractiveness was appraised in a glance (20ms). The phenomenon demonstrates the impact of the interaction between processing of the visible boundary and the internal features of the face on appraisal of facial beauty. This rapid holistic integration effect is striking because the shape of a mask is not a part of the face, but of an occluding object, although it is apparently assigned to the face.

ABSTRACT. Conventional thinking about how pupil reflects attractiveness predicts that pupil dilated when seeing attractive faces. Contradictorily, in the current study, we showed that pupil constricts, instead of dilates, to attractive faces. Participants looked at a face presented at the center of the screen, rated how attractive the face was from 1 to 9, while an infrared camera recorded their pupillary responses. Results showed that participants made their decision within 2 s from the stimulus onset, during which pupil constricted more for the more attractive (as-to-be-rated) faces. Control experiments showed that the pupil constriction was limited to the attractiveness task (thus not in the roundness task) and to faces (thus not in the attractiveness task on geometric shapes). To examine whether pupil constriction (caused by other factor) affects facial attractiveness, we manipulated the luminance of the pre-target display and the background luminance of the target display, to induce different degrees of pupil constriction. Results showed that faces were judged to be more attractive following a black display (where pupil constricted more) than a gray display (where pupil constricted less), regardless of the background luminance of the target display (where pupil constriction did not differ much). In a questionnaire after the experiment, most participants reported awareness of the background luminance change but did not have a correct idea of how the luminance parameters affect their attractiveness ratings. The findings are discussed in relation to the parasympathetic system due to appraisal and enjoy, and the positive loop of seeing and liking (Shimojo et al., 2003). In sum, the overall results suggest that pupil constriction not only reflects but also affects facial attractiveness implicitly.

ABSTRACT. Adaptation reflects the ability of the human visual system to adjust and fine-tune to a constantly changing environment. This study seeks to investigate the adaptation aftereffects in facial attractiveness. Previous study (Hayn-Leichsenring, Kloth, Schweinberger, & Redies, 2013) on face attractiveness adaptation used different faces based on their ratings of attractiveness, which was a discrete measurement, and the same sets of faces were used for all subjects. Hence, it lacked a continuum in judgement of face attractiveness and ignored individual differences in face attractiveness perception. On the contrary, in the first task, we asked each participant to rate a series of faces on attractiveness. This was done to ensure that the two extreme ends of the facial attractiveness continuum were indeed perceived to be highly attractive or highly unattractive to each participant. Face morphs with varying degrees of attractiveness in this continuum were then generated. In the second task, participants adapted to attractive/unattractive male/female faces. There were six conditions in total, adapting to: attractive female faces, unattractive female faces, attractive male faces, and unattractive male faces, and two baseline conditions without adaptation. Results from paired-samples t tests revealed a significant effect for all conditions relative to their respective baselines, such as after adapting to an attractive female face, a subsequent female face was more likely to be judged as unattractive. Likewise, after adapting to an unattractive female face, a subsequent female face was more likely to be judged as attractive. A similar pattern was observed for the male faces. These results suggest that our perception of facial attractiveness can be biased by prior experiences. Such effect has important social implications, and the mechanisms of this effect remain to be further explored.

ABSTRACT. Many classic experiments suggest that the face is processed as a whole configuration rather than feature by feature. If we represent faces based upon their configural information, we should be good at memorizing configurations for later recognition. However, surprisingly little research has addressed the question of “how well we can recall the configuration of a face”. Here we propose a new methodology – a face puzzle task to test memory of a face configuration. Participants were asked to put a set of facial features (two eyebrows, two eyes, nose and mouth) back into correct position in order to make a face. Participants were randomly assigned into two groups. One group studied 40 upright faces before performing the face puzzle task in either upright or inverted orientations. The overall accuracy and the configural difference, such as the inter-pupil distance, eye-to-mouth distance and so on, of the reassembled face was recorded and compared with the original faces. Another group of participants served as a control group; they did the face puzzle task on the same 40 faces without any prior study. The result found no difference between the study and non-study groups. Both groups showed an inversion effect, where they performed better for upright faces than inverted faces. This result demonstrates that configural information is not used by people to memorize and recall a face is. This is consistent with a recent view that challenges the role of configural processing in face recognition.

ABSTRACT. Familiarity preference and novelty preference co-exist in our daily decision making, and a systematic understanding of when these two seemingly-opposite preferences exhibit is desirable. Park, Shimojo, and Shimojo (2010) reported that object category matters for adults’ preference formation: we prefer familiar faces while favor novel nature scenes. We explored whether this distinction was already present during infancy.

We recruited 8-10 months old Chinese infants and exposed them to four object categories: mountains, animals, female and male faces. In each trial, an old/familiar image was always the same and placed side-by-side to a new/novel image from the same category. Infants’ gazes were monitored and recorded with a Tobii eye-tracker. For comparison, we recruited adults to rate their preference and record their free viewing eye gaze.

For the nature scene condition, infants looked significantly longer to novel images, consistent to the adults. For faces, we found that infants had no visual preference toward female faces but a novelty preference toward male faces. We did not observe gender effect in adults’ rating: they prefer familiar faces similar to the study in Park et al (2010).

In summary, infants’ novelty preference dominated across face and non-face categories except female faces, possibly the most significantly and frequently viewed objects in human early life. Further studies are called to examine this distinction and other factors modulating infants’ preference formation. We did not find strong evidence for face- v.s. non-face categories that modulated 8-10 month-old infants’ viewing preference, which suggests this categorical distinction comes later in life.

ABSTRACT. Figure-ground segregation is a crucial step toward object recognition. Physiological studies have reported that a number of neurons in V2-V4 of Macaque monkeys are selective to the local direction of figural region along a contour (Border Ownership)[e.g., Zhou et al., J.Neurosci, 2000]. We have reported that asymmetric surround modulation reproduces the characteristics of BO-selective cells [e.g., Sakai et al., J.Cogn.Neurosci, 2006]. However, the determination of BO along complex contours in natural images appears difficult for a single BO-selective cell, leading to an idea that multiple cells work simultaneously as a population for yielding the veridical perception of BO. We investigated whether a group of cells represents BO along a local contour in natural images. Specifically, we generated a set of various local contours from natural scenes in BSDS [Martin et al., ICCV, 2001], and presented to the BO models with distinct surround regions. Because the neural mechanism for integration has not been clarified, we applied a versatile machine-learning technique (Support Vector Machine) to the outputs of individual model cells so as to maximize the discrimination rate. This procedure is equivalent to the computation of theoretical upper limit given the outputs of the model cells. The results of simulations showed that 10 model-cells yield around 80% correct for the local contours from natural scenes, suggesting that grouping a small number of BO-selective cells increases effectively the capability for the determination of BO. We also applied the SVM to the responses of 10-20 neurons recorded electrophysiologically from V4 of 2 macaque monkeys (Macaca fuscata). We observed a discrimination rate that was significant but slightly lower than that of the model. These results suggest that a small number, several to a few tens, of cortical cells with simple mechanisms are capable of determining the direction of figure from a variety of natural contours.

ABSTRACT. We use computational and experimental approaches to gain insights into visual and cognitive functions and disorders, and to build intelligent visual systems that approach human performance. 
In this talk, I will introduce our recent efforts in understanding and predicting visual attention in natural scenes. I will first share new perspectives and observations in selective attention in natural scene images, with a new attention architecture that incorporates information at multiple layers: low level, object level, and semantic level, and a new natural scene dataset with well-characterized object and semantic information. I will also present an innovative psychophysical method to allow large-scale collection of “eye tracking” data using mouse or touch pad / screen, thus enabling crowd sourcing the task. Eye tracking is commonly used in visual neuroscience and cognitive science to study questions in visual attention and decision making. We envision that bigger eye tracking data will advance the understanding of these questions due to the inherently complex nature of both the natural scene stimuli and human cognitive processes. Insights derived from human data will be highlighted and live demos will be shown.

ABSTRACT. In this talk, we present two recent studies trying to find the neural cause of human visual perception and behavior. One of the promising techniques for studying causality is fMRI decoded neurofeedback (DecNef), which can non-invasively induce neural activities corresponding to specific information (e.g. color and motion). Our previous study has shown that an achromatic grating was perceived to be reddish after the induction of V1/V2 activity pattern corresponding to red color during the presentation of the achromatic grating. In the current study, we first built a decoder of confidence rating during direction judgment of random-dot coherent motion. Confidence was found to be decoded from activity patterns in frontoparietal areas. Subjects then implicitly learned to induce activity patterns corresponding to high or low confidence in these areas. After this training, confidence in direction judgment was increased or decreased while the task performance was kept constant. These results suggest that frontoparietal areas might be causally related to the confidence in perceptual judgment.
In the second part, we present an MEG experiment showing a possible functional role of alpha oscillation in visual processing. When borders defined by iso-luminant color change and those defined by luminance change are placed in close proximity, the color-defined border is perceived to be jittering at around 10 Hz (Arnold and Johnston, 2003; Amano et al., 2008). We have recently found that the individual alpha frequency during rest is highly correlated with the perceived jitter frequency, suggesting the possibility that intrinsic alpha rhythm could be causally related to visual processing, as a compensation for a spatial delay of the iso-luminant border originating from its slower perceived speed in this case. We propose the hypothesis that dorsal and ventral visual areas communicate at the alpha rhythm to resolve the dissociation between motion-based and object-based position representations.

ABSTRACT. Vision has been said to be "knowing what is where by seeing". This view is supported, in part, by the rough separation of the primate visual system into "what" and "where" streams. Meanwhile, in computer vision, many problems can be similarly divided into "what" problems (e.g. object categorization), "where" problems (e.g. segmentation) and their interaction (e.g. tracking). Determining “what is where” in an image is a fundamental part of vision. However, I will argue that people possess visual intelligence that is capable of much more than simply knowing what is where. In this talk, I will propose ideas about what problems, tasks, or routines might compose the notion of visual intelligence. I will then examine these problems, tasks and routines from the perspectives of neuroscience, psychology and computer science, discussing where the research gaps are, and whether it will fruitful to try to fill these gaps.

ABSTRACT. Deep convolutional neural networks (DCNNs), which is inspired from the vision system, show good performances for the recognition task in the field of computer vision and machine learning in these years. DCNN is a kind of multi layer neural network that can learn feature representation including in the input by use of large amount of training data. However, in the several fields such like medical imaging, the acquisition cost of large amount of data is high since it requires diagnosed labels. The small number of the training data for the DCNN sometimes occurs overt-raining. So that, the large scale DCNN looks hard to apply such kind of small dataset directly. On the contrary, human physicians can adapt such recognition task without overt-raining. In this study, we introduced the DCNN, which is trained with large amount of natural image data, for the initial state of the novel medical image recognition task. The recognition task is to classify the diffuse lung diseases (DLD) on High-resolution computed tomography (HRCT). At first, we apply massive natural images to train the DCNN for the pre-training. After that, we train the trained DCNN with the small number of the DLD HRCT images as the labeled data for fine-tuning. Thus, we can compare DCNNs with training of 1) DLD HRCT images only, 2) natural images only, and 3) DLD HRCT images + natural images. In the result, we show the results of 3) performs 80.04% accuracy, which is the best classification results among those features. Moreover, the results of 2), that performs 75.78%, is just better than that of the case 1) that is 74.13%. Thus, we conclude the pre-matured initial state is important for such novel pattern classification.

ABSTRACT. Recent studies on image memorability have shed light on what distinguishes the memorability of different images and the intrinsic and extrinsic properties that make those images memorable. However, a clear understanding of the memorability of specific objects inside an image remains elusive. In this paper, we provide the first attempt to answer the question: what exactly about an image is remembered? We augment both the images and object segmentations from a popular object recognition dataset with ground truth memorability scores and shed light on the various factors and properties that make an object memorable (or forgettable) to humans. We analyze various visual factors that may influence object memorability (e.g. color, visual saliency, and object categories). We show that the category of an object has meaningful influence on its memorability, and that visual saliency can predict object memorability to some degree. We also study the correlation between object and image memorability and find that image memorability is greatly affected by the memorability of its most memorable object. Lastly, we explore the effectiveness of deep learning and other computational approaches in predicting object memorability in images. We find that a simple baseline model that utilizes a convolutional neural network trained on the ImageNet database performs favorably against state-of-the-art computational saliency models. Our efforts offer a deeper understanding of memorability in general thereby opening up avenues for a wide variety of applications.

ABSTRACT. The present study tested the dominant notion that processing of familiar face takes place in an automatic, capacity-unlimited manner. To do so, we adopted a well-established paradigm to reveal the capacity-limited nature of perception: the simultaneous-sequential paradigm. Specifically, we had a group of participants search for their own face among others’ faces. Importantly, the target and distractors were presented in two different ways. In the simultaneous presentation condition, all the search stimuli were briefly presented at the same time. In the sequential presentation condition, only a half of the stimuli were briefly presented first. Then, after a certain period of interval, the other half was presented. To examine whether this self-face processing is subject to capacity limit, we compared the search performance between when all the search stimuli were simultaneously presented and when two different subsets of search stimuli were sequentially presented. If the process of detecting self-face suffers from capacity limit, the search performance should be better in the sequential presentation than in the simultaneous presentation condition. By contrast, if one can detect her/his own face in an automatic, capacity-unlimited manner, there should be no performance difference across different presentation conditions. The results showed that the search performance (perceptual sensitivity to the target, d prime) was benefitted from the sequential presentation, t(14) = 3.62, p < .01, indicating that detecting one’s own face depends on a capacity-limited attentional process. A similar pattern of result was found when participants searched for someone else’s face, t(14) = 3.64, p < .005, although there was decline in the overall performance, p < .05. These findings suggest that the processes of detecting familiar and unfamiliar faces suffer from capacity limit of visual perception to a similar extent, challenging the notion of automaticity of familiar face processing.

ABSTRACT. Inhibition of return (IOR) is a robust cognitive phenomenon whereby manual and oculomotor responses to target stimuli are slowed when the target location has been previously attended. IOR is thought to subserve novelty seeking and foraging facilitation in the real world. Two different forms of IOR have been identified – an input-based form when the oculomotor system is actively suppressed, and an output-based form when eye movements are allowed. The vast majority of previous work investigating IOR with event-related potentials (ERPs) has focused on the input-based form of IOR, given that eye movements are normally suppressed in ERP studies. In these studies, modulations of the early sensory P1 ERP component have been associated with behavioural IOR. However, recent work has demonstrated that when the eye movement system is activated – thereby initiating output-based, oculomotor IOR – P1 reductions are not correlated with IOR. Output-based IOR is present, and equivalent, regardless of whether or not stimuli are presented as exogenous peripheral stimulations or endogenous central arrows, however, P1 reductions are only observed with peripheral stimuli. Furthermore, when retinotopic and spatiotopic coordinate systems are dissociated, P1 reductions are only found in the retinotopic condition, whereas oculomotor IOR is only found in the spatiotopic condition. The later Nd ERP component, however, is present in association with IOR in conditions using central arrows and spatiotopic coordinate systems. We therefore propose that P1 reductions are simply a result of repeated peripheral stimulation at the same location and they do not reflect oculomotor IOR. Later components, such as the Nd and possibly N2pc components, are more likely to reflect a neurophysiological measurement of true, oculomotor IOR.

ABSTRACT. Attention and awareness are two tightly coupled processes that have been the subject of the same enduring debate: Are they allocated in a discrete or in a graded fashion? Using the attentional blink paradigm with a psychophysics approach, we show that awareness arises at central stages of information processing in an all-or-none manner. Manipulating the temporal delay between two targets affected subjects’ likelihood of consciously perceiving the second target, but did not affect relative contrast judgments between the two targets. Conversely, when the same stimuli were used in a spatial attention cueing paradigm (a replication of Carrasco, Ling, & Reid, 2004), cueing did affect relative contrast judgments. The findings distinguish the fundamental contributions of attention and awareness at central stages of visual cognition: Conscious perception emerges in a quantal manner, with attention serving to modulate the probability that representations reach awareness.

ABSTRACT. Previous experiments have shown that people miss a visible but unexpected object in a field of other objects when engaged in an attentionally demanding task. Yet no previous experiment has examined the extreme case of blindness to a single object appearing alone in the visual field. In two experiments we evaluated whether a perceiver’s prior expectations could alone obliterate his or her awareness of a salient visual stimulus. To establish expectancy, observers first made a demanding visual discrimination on each of three baseline trials. Then, on a fourth, critical trial, a single, salient and highly visible object appeared in full view at the center of the visual field and in the absence of any competing visual input or competing task demands. Surprisingly, fully half of the participants were unaware of the solitary object in front of their eyes. Yet in a control condition these same observers easily detected the same object in the same screen position. Dramatically, observers were even blind when the only stimulus on display was the face of U.S. President Barack Obama. We term this novel, counterintuitive phenomenon, Barack Obama Blindness (BOB). Employing a method that rules out putative memory effects by probing awareness immediately after presentation of the critical stimulus, we demonstrate that the BOB effect is a true failure of conscious vision.

ABSTRACT. This research aims to investigate whether attention can modulate the formation of long-distance, pattern-dependent contextual effect.

In experiment 1, we varied participants’ attentional resource via changing task requirement while recording the EEG signal. For one group, we asked them to finish a contrast changes detection task on fixation dot. For another group of subjects, we asked them to passively view the fixation dot. In both tasks, participants were asked to ignore the changes of the Gabor patches. Thus we manipulated how much attention was deployed to the surrounding Gabor patches.

Results show that the long-distance, pattern-dependent contextual effect appears when there is fixation task requirement, Gabor patches in ‘Orthogonal’ condition elicit larger amplitude than the other two conditions. This result is consistent with previous study. However, there is no significant difference on P1 component among three orientation configurations when participants have to finish the required task, such long-distance, pattern-dependent contextual effect disappeared.

However, the findings in experiment 1 can be also explained as the results of different task requirements rather than the attentional modulation. We develop experiment 2 so as to roll out such alternative explanation. In experiment 2, we used the attentional load paradigm to control attentional level while most of other settings were same as in experiment 1. Results show a significant behavioral difference on accuracy and response time and also a remarkable amplitude difference on C1 component in low attentional load condition. However such effect disappears in high attentional load condition. The observed results in experiment 2 confirm our findings in experiment 1.

These findings in two experiments confirm the long-distance, pattern- dependent contextual effect in human early visual cortex. The results also show a significant interaction between attention and long-distance, pattern-dependent contextual effect. Such findings extend our understanding of how attention interacts with early visual processing.