- Email: [email protected]
Visual experience modifies the neural dynamics for orientation processing in cat visual cortex
S108
Abstracts
behavior. In contrast, Nr neurons did not exhibit such selectivity nor correlation. These results suggest that different classes of parietal neurons contribute differently to the visual grouping of discrete elements. doi:10.1016/j.neures.2009.09.484
P1-h03 Contribution of color signals to ocular following responses Kiyoto Matsuura, Kenichiro Miura, Kenji Kawano Grad. Schl. Med., Kyoto University, Kyoto, Japan Ocular following responses (OFRs) are elicited with ultra-short latencies (∼60 ms) by sudden motion of a visual scene. It is known that the OFRs are largely influenced by luminance signals which are mediated by the magno system. However, little is known whether the OFRs are influenced by color signals which are processed by the parvo system. First, we made vertical isoluminant color sinusoidal gratings with minimum motion technique and recorded the OFRs to horizontal motion of the gratings in monkeys. We found that the isoluminant color gratings did elicit the OFRs. Then, we recorded the OFRs to moving color-gratings and luminance-gratings at 6 spatial (0.07–2.3 cpd) and 5 temporal frequencies (1.56–25 Hz). We found that the high spatial and low temporal frequencies were more suitable for the color-gratings to elicit the OFRs than for the luminance-gratings. These results suggest that the OFRs can also be driven by the signals mediated by the parvo system whose visual responses show high spatial and low temporal resolution. doi:10.1016/j.neures.2009.09.485
P1-h04 Spatial resolution of direction discrimination: comparison of MT neurons and behavior Hironori Kumano, Takanori Uka Dept. Physiol 1, Juntendo University School of Medicine, Japan Observers have difficulty in identifying a target in the presence of surrounding distracters. To understand the neuronal mechanisms underlying this phenomenon, known as ‘crowding’, we recorded activities of single MT neurons from a monkey performing direction discrimination in a bipartite center/surround motion display. The monkey was required to discriminate direction of the center patch of moving random dots, of which parameters were tailored to selectivity of the neuron under study, and to ignore the surrounding noise of varying diameter. Psychophysical threshold first increased then decreased as the noise diameter increased, suggesting an improvement of spatial resolution of direction discrimination at the large noise condition (‘anti-crowding’). Neuronal threshold of MT neurons also first increased then decreased with increasing surround size. The increased sensitivity with a large surround is due to increased differences of mean responses, not reduced variability. The results suggest that changes of sensitivities in MT neurons underlie the changes of sensitivities observed psychophysically. doi:10.1016/j.neures.2009.09.486
P1-h06 Automatic extraction of visual image bases from fMRI response patterns Yusuke Fujiwara 1 , Yoichi Miyawaki 1,2 , Yukiyasu Kamitani 1,3 1 3
ATR Computational Neuroscience Laboratories, Japan; NAIST, Japan
IPM School of Cognitive Sciences, Tehran, Iran Here we examined monkeys’ performance in categorizing ambiguous visual objects. We trained monkeys to perform a two-alternative forced-choice body/object categorization task. Stimuli were presented in 5 noise levels. After monkey fixated on a monitor for 400 ms, a noisy image was presented for 70 ms, followed by 500 ms blank interval and then two response targets in the left and right of the screen centre, indicating body and object responses, respectively. Monkeys had to make a saccade to the correct target no later than 300 ms after target onset. Behaviorald’ analysis showed a linear decline of performance as noisier stimuli were presented. But for a given noise level there was a difference in the amount of performance decline between body and object. Also there was a correlation between performance and reaction time. Although there was a 500 ms time delay between stimuli and targets, reaction time increased significantly as noisier images were presented. Mean difference between reaction time of the least and most noisy images was about 30 ms. These results advance our understanding of the nature of visual object categorization in non-human primates. doi:10.1016/j.neures.2009.09.487
NICT, Japan;
Image bases provide a framework for understanding neural representation of visual perception. We have recently shown that arbitrary contrast-defined visual images can be reconstructed from fMRI activity patterns of early visual cortex, by a combination of multi-scale local image bases (Miyawaki et al., Neuron 2008). Our model assumed fixed multi-scale image bases, whose contrasts were estimated from fMRI activity patterns. Such heuristically determined image bases may not be optimal for reconstruction. Here, we propose a method for automatically extracting image bases from fMRI data. We constructed a probabilistic model that relates the fMRI activity space to the visual image space via an intermediate representation. The mapping from the intermediate representation to the visual image space can be regarded as a set of image bases. The proposed model was trained with fMRI data measured while a subject viewed random images. As a result, spatially localized image bases were obtained around the foveal region and able to reconstruct novel visual images. doi:10.1016/j.neures.2009.09.488
P1-h07 Understanding brain information processing in the visual system following flickering stimulation Monique Maurice 1 , Vialatte Francois-Benoit 2 , Bakardjian Hovagim 2 , Cichocki Andrzej 2 , Yamaguchi Yoko 1 1 Dynamics of Emergent Intelligence, RIKEN Brain Science Institute, Wako, Japan; 2 Laboratory for Advanced Brain Signal Processing, RIKEN Brain Science Institute, Wako, Japan
Our main purpose is to understand brain information processing in the visual system following to flickering stimulation. In order to do so, subjects were stimulated during fMRI and EEG experimentation. During stimulation, the subject sees a flickering white/black light, displayed using Silent Vision goggles (visual angle 18 × 24 degrees), generated with a Shutter to control high frequencies. During fixation the subject sees a gray background, isoluminant with the stimulus. Experiments are performed for 21 different frequencies from 1 to 100 Hz. Frequencies are pseudorandomized and the same sequence of frequencies is presented to all subjects. We report specific effects (location and dynamics) for low, medium, and high frequencies; and particularly investigate the outcome of flickering frequencies in very low (<5 Hz) and very high (>40 Hz) subbands doi:10.1016/j.neures.2009.09.489
P1-h08 Receptive field extraction in free viewing condition Michel Vidal-Naquet 1 , Pedro Maldonado 2 , Sonja Gruen 1 1
P1-h05 Categorization of ambiguous visual objects by macaque monkeys Nazli Emadi, Hossein Esteky
2
RIKEN Brain Science Institute, Japan; Chile
2
Universidad de Chile, Santiago,
Electrophysiological multi-neuron recordings in free viewing monkeys are limited in time and therefore make a reliable extraction of visual receptive fields (RF) difficult, since this normally requires a large set of well controlled stimuli (∼10,000). However, knowledge of the RFs is highly relevant in order to relate neuronal responses to behavior. Here, we propose a method inspired by [Vidal-Naquet et al., COSYNE’09] to extract V1 receptive fields in conditions of freely viewing natural scenes, when the number of stimuli observed by the subject is significantly smaller (∼100). Using simulations, we show that our method can recover the structure and spatial location of noisy, non-linear receptive fields from the recorded saccade point locations on observed natural stimuli together with the related neural activity. Our method relies on the use of a statistical test that can reliably predict whether the estimated receptive field is significant or not. Finally, we present receptive fields extracted using our method from real physiological recordings. doi:10.1016/j.neures.2009.09.490
P1-h09 Visual experience modifies the neural dynamics for orientation processing in cat visual cortex Kazunori O’Hashi 1,2 , Amiram Grinvald 1 , Shigeru Tanaka 2 1
Weizmann Institute, Israel;
2
RIKEN BSI, Japan
Orientation selectivity of visual cortex is highly adaptive to visual experience in early postnatal life. Orientation-restricted environment during the critical period can transform the orientation representation to respond to the exposed orientation preferably. It remains unclear, however, how orientation signals are processed by the altered visual cortical circuits. In this study, we examined spatiotemporal dynamics
Abstracts for orientation processing in the visual cortex of the goggle-reared animals using voltage-sensitive dye imaging. As a result, we found that the response onset to the exposed orientation was faster than the others, whereas the responses to unexposed orientations showed brief but strong hyperpolarization followed by depolarization. In addition, orientation preferences were temporally unstable at late period of visual stimulation. After 1-week normal viewing, however, the temporal dynamics of visual responses were restored to those in normally reared animals. These findings suggest that the visual cortical circuits change the way of information processing depending on visual experience. doi:10.1016/j.neures.2009.09.491
P1-h10 Neural correlates of intersubject variability of continuous flash suppression Hiroyuki Yamashiro 1,2 , Hiroki Yamamoto 1 , Jun Saiki 1 , Hiroaki Mano 1,4 , Masahiro Umeda 3 , Chuzo Tanaka 4 1
Grad. Schl. of Human & Environmental Stud., Kyoto Univ, Kyoto, Japan; JSPS, Tokyo, Japan; 3 Department of Med. Informatics, Meiji University of Integrative Med., Kyoto, Japan; 4 Department of Neurosurgery, Meiji University of Integrative Med., Kyoto, Japan
2
In continuous flash suppression (CFS), a monocular target stimulus can be suppressed from visual awareness for up to several minutes by a presence of a highly dynamic mask in the other eye. Interestingly, the durations of suppression depend greatly on subjects, which range from several seconds to minutes. To explore the neural correlates of the intersubject variability of CFS, we measured brain activity using fMRI, while subjects viewed a rotating target rendered invisible by CFS. We found a strong correlation between the effects of CFS and the retinotopic responses to the invisible target in extrastriate visual areas (V3, V3A, V4, V7, MT, LOc): the less susceptible subjects were to CFS, the larger responses were evoked. This suggests that extrastriate visual areas play a critical role in generating and suppressing visual awareness. doi:10.1016/j.neures.2009.09.492
P1-h11 Monkeys perceive orientations relative to vertical Masumi Wakita Primate Res Inst, Kyoto Univ, Aichi, Japan Monkeys readily discriminate cardinal orientations (horizontal and vertical) from oblique orientations. In addition, they recognize cardinals without confusing with any other orientations. Intuitively, cardinals may be a reference to judge degree or direction of obliques. However, functional difference between horizontal and vertical remains unclear. Here, five monkeys underwent two discrimination training between 22.5◦ and 157.5◦ gratings (horizontal condition; stimulus separation was 45◦ across horizontal axis), and between 67.5◦ and 112.5◦ gratings (vertical condition; stimulus separation was 45◦ across vertical axis). After each training session, their performance was tested using gratings from 0◦ to 157.5◦ with a step of 22.5◦ . Consequently, generalization tests revealed that animals showed wider generalization curve in the vertical condition compared with the horizontal condition. In other words, monkeys perceived a distance between 67.5◦ and 112.5◦ was smaller than that between 22.5◦ and 157.5◦ . Thus, they perceive orientations relative to vertical axis. doi:10.1016/j.neures.2009.09.493
P1-h12 Higher-order structures in natural scenes Takuma Tanaka 1 , Toshio Aoyagi 2 , Takeshi Kaneko 1 1 2
Department of Morphogical Brain Science, Kyoto Univ, Kyoto, Japan; Graduate School of Informatics, Kyoto University, Kyoto, Japan
It has been shown that the selectivity of neurons in lateral geniculate nucleus and simple cells in the primary visual cortex can be reproduced by the models which optimize the information transmission between the input and output units. Recently, Karklin and Lewicki (2009) reported that complex cell-like selectivity can be selforganized through the learning of the probability distribution of natural images. We present a model in which the activity of neurons represents the probability of the present input. Using the output of simple cell-like linear filters as the input, we found that neurons in this model after learning exhibited the orientation preference and translation invariance, both of which are properties of complex cells. We conclude that complex cells detect and represent the higher-order structures of the probability distribution of natural scenes. doi:10.1016/j.neures.2009.09.494
S109
P1-h13 Transfer of synesthetic experience Aleksandra Mroczko 1,2 , Thomas Metzinger 2,3 , Wolf Singer 1,3 , Danko Nikolic 1,3 1
Max Planck Institute for Brain Research, Germany; 2 University of Mainz, Germany; 3 Frankfurt Institute for Advanced Studies, Germany
In synesthesia, a certain stimulus (e.g. grapheme) is associated automatically and consistently with a stable perceptual-like experience (e.g. color). We found that grapheme-color associations can transfer to novel graphemes after only a 10-minute writing exercise. Most subjects experienced synesthetic associations immediately after learning a new Glagolitic grapheme. Using a Stroop task, we provide objective evidence for the creation of novel associations between the newly learned graphemes and synesthetic colors. Also, these associations generalized to graphemes handwritten by another person. The fast learning process and the generalization suggest that synesthesis begins at the semantic level of representation with the activation of a certain concept (the inducer), which then, uniquely for the synesthetes, activates representations at the perceptual level (the concurrent). doi:10.1016/j.neures.2009.09.495
P1-h14 Target-specific connectivity enhances the stability of activity dynamics in a layered cortical network model Tobias C. Potjans 1,2 , Markus Diesmann 2,3,4 1
2 Research Center Juelich, Germany; RIKEN Computational Science Research Program, Wako-shi, Japan; 3 RIKEN Brain Science Institute, Wakoshi, Japan; 4 Bernstein Center for Computational Neuroscience, Freiburg, Germany
The local cortical network consists of distinctively interconnected layers. In addition to the well-established feedforward pattern of connections (L4 to L2/3 to L5 to L6), specific feedback connections have been identified that specifically select interneuronal targets. We employ large-scale simulations to investigate the dynamical consequences of these target specific connections. Our model consists of 80,000 neurons and reflects the layer-specific connectivity established in anatomical and physiological studies. We find that target-specific connections enhance the stability of network activity. Furthermore, target specificity increases the sensitivity for time-dependent signaling: only these networks exhibit strong, temporally locked responses to transient thalamic inputs. doi:10.1016/j.neures.2009.09.496
P1-h15 A gaze prediction model considering the inhibition of return based on the saliency map theory Taiki Gochi, Takeshi Kohama Grad. Sch. Biology-Oriented Science and Technology, University of Kinki, Wakayama Japan Previous saliency-based visual attention models deal with the mechanism of attentional shifts, and such models can predict the distribution of gaze movements based on the saliency map. Nevertheless, repeating a constant movement pattern in the prediction model poses a problem. Therefore, in order to improve this unrealistic prediction, it is necessary to consider the abstraction of the inhibition of return (IOR). IOR refers to the suppression of the processing of objects that have recently been gazed at attentively and prevents the viewer from watching the same objects repeatedly. In previous models, IOR operates at constant intensity in the higher stage of processing; however, the intensity of inhibition must depend on the importance of the gazed object. In this study, we modified the IOR operation by the Gaussian proportion decrement with random peak values in the lower-stage processing. As a result, the predictions of gaze distribution were more sparse and realistic. This indicates that the inhibition of return might be a low-order operation at variant intensity. doi:10.1016/j.neures.2009.09.497
P1-h16 The influence of figure components and their motion direction on motion position perception Azusa Deguchi, Akitoshi Hanazawa Department Brain Sci. Eng., Kyusyu Institute of Technology, Kitakyusyu, Japan We found a new optical illusion. We made a figure consisting of one long line and many short lines, which are perpendicular to it, and turned this figure around the middle point of the long line. Then rotating position of the short lines was perceived behind the long line. This phenomenon can be thought as follows. For the short line, because motion direction is parallel to the axis of the line, motion is detected only at the end points. On the other hand, for the long line, because motion direction is
Abstracts
behavior. In contrast, Nr neurons did not exhibit such selectivity nor correlation. These results suggest that different classes of parietal neurons contribute differently to the visual grouping of discrete elements. doi:10.1016/j.neures.2009.09.484
P1-h03 Contribution of color signals to ocular following responses Kiyoto Matsuura, Kenichiro Miura, Kenji Kawano Grad. Schl. Med., Kyoto University, Kyoto, Japan Ocular following responses (OFRs) are elicited with ultra-short latencies (∼60 ms) by sudden motion of a visual scene. It is known that the OFRs are largely influenced by luminance signals which are mediated by the magno system. However, little is known whether the OFRs are influenced by color signals which are processed by the parvo system. First, we made vertical isoluminant color sinusoidal gratings with minimum motion technique and recorded the OFRs to horizontal motion of the gratings in monkeys. We found that the isoluminant color gratings did elicit the OFRs. Then, we recorded the OFRs to moving color-gratings and luminance-gratings at 6 spatial (0.07–2.3 cpd) and 5 temporal frequencies (1.56–25 Hz). We found that the high spatial and low temporal frequencies were more suitable for the color-gratings to elicit the OFRs than for the luminance-gratings. These results suggest that the OFRs can also be driven by the signals mediated by the parvo system whose visual responses show high spatial and low temporal resolution. doi:10.1016/j.neures.2009.09.485
P1-h04 Spatial resolution of direction discrimination: comparison of MT neurons and behavior Hironori Kumano, Takanori Uka Dept. Physiol 1, Juntendo University School of Medicine, Japan Observers have difficulty in identifying a target in the presence of surrounding distracters. To understand the neuronal mechanisms underlying this phenomenon, known as ‘crowding’, we recorded activities of single MT neurons from a monkey performing direction discrimination in a bipartite center/surround motion display. The monkey was required to discriminate direction of the center patch of moving random dots, of which parameters were tailored to selectivity of the neuron under study, and to ignore the surrounding noise of varying diameter. Psychophysical threshold first increased then decreased as the noise diameter increased, suggesting an improvement of spatial resolution of direction discrimination at the large noise condition (‘anti-crowding’). Neuronal threshold of MT neurons also first increased then decreased with increasing surround size. The increased sensitivity with a large surround is due to increased differences of mean responses, not reduced variability. The results suggest that changes of sensitivities in MT neurons underlie the changes of sensitivities observed psychophysically. doi:10.1016/j.neures.2009.09.486
P1-h06 Automatic extraction of visual image bases from fMRI response patterns Yusuke Fujiwara 1 , Yoichi Miyawaki 1,2 , Yukiyasu Kamitani 1,3 1 3
ATR Computational Neuroscience Laboratories, Japan; NAIST, Japan
IPM School of Cognitive Sciences, Tehran, Iran Here we examined monkeys’ performance in categorizing ambiguous visual objects. We trained monkeys to perform a two-alternative forced-choice body/object categorization task. Stimuli were presented in 5 noise levels. After monkey fixated on a monitor for 400 ms, a noisy image was presented for 70 ms, followed by 500 ms blank interval and then two response targets in the left and right of the screen centre, indicating body and object responses, respectively. Monkeys had to make a saccade to the correct target no later than 300 ms after target onset. Behaviorald’ analysis showed a linear decline of performance as noisier stimuli were presented. But for a given noise level there was a difference in the amount of performance decline between body and object. Also there was a correlation between performance and reaction time. Although there was a 500 ms time delay between stimuli and targets, reaction time increased significantly as noisier images were presented. Mean difference between reaction time of the least and most noisy images was about 30 ms. These results advance our understanding of the nature of visual object categorization in non-human primates. doi:10.1016/j.neures.2009.09.487
NICT, Japan;
Image bases provide a framework for understanding neural representation of visual perception. We have recently shown that arbitrary contrast-defined visual images can be reconstructed from fMRI activity patterns of early visual cortex, by a combination of multi-scale local image bases (Miyawaki et al., Neuron 2008). Our model assumed fixed multi-scale image bases, whose contrasts were estimated from fMRI activity patterns. Such heuristically determined image bases may not be optimal for reconstruction. Here, we propose a method for automatically extracting image bases from fMRI data. We constructed a probabilistic model that relates the fMRI activity space to the visual image space via an intermediate representation. The mapping from the intermediate representation to the visual image space can be regarded as a set of image bases. The proposed model was trained with fMRI data measured while a subject viewed random images. As a result, spatially localized image bases were obtained around the foveal region and able to reconstruct novel visual images. doi:10.1016/j.neures.2009.09.488
P1-h07 Understanding brain information processing in the visual system following flickering stimulation Monique Maurice 1 , Vialatte Francois-Benoit 2 , Bakardjian Hovagim 2 , Cichocki Andrzej 2 , Yamaguchi Yoko 1 1 Dynamics of Emergent Intelligence, RIKEN Brain Science Institute, Wako, Japan; 2 Laboratory for Advanced Brain Signal Processing, RIKEN Brain Science Institute, Wako, Japan
Our main purpose is to understand brain information processing in the visual system following to flickering stimulation. In order to do so, subjects were stimulated during fMRI and EEG experimentation. During stimulation, the subject sees a flickering white/black light, displayed using Silent Vision goggles (visual angle 18 × 24 degrees), generated with a Shutter to control high frequencies. During fixation the subject sees a gray background, isoluminant with the stimulus. Experiments are performed for 21 different frequencies from 1 to 100 Hz. Frequencies are pseudorandomized and the same sequence of frequencies is presented to all subjects. We report specific effects (location and dynamics) for low, medium, and high frequencies; and particularly investigate the outcome of flickering frequencies in very low (<5 Hz) and very high (>40 Hz) subbands doi:10.1016/j.neures.2009.09.489
P1-h08 Receptive field extraction in free viewing condition Michel Vidal-Naquet 1 , Pedro Maldonado 2 , Sonja Gruen 1 1
P1-h05 Categorization of ambiguous visual objects by macaque monkeys Nazli Emadi, Hossein Esteky
2
RIKEN Brain Science Institute, Japan; Chile
2
Universidad de Chile, Santiago,
Electrophysiological multi-neuron recordings in free viewing monkeys are limited in time and therefore make a reliable extraction of visual receptive fields (RF) difficult, since this normally requires a large set of well controlled stimuli (∼10,000). However, knowledge of the RFs is highly relevant in order to relate neuronal responses to behavior. Here, we propose a method inspired by [Vidal-Naquet et al., COSYNE’09] to extract V1 receptive fields in conditions of freely viewing natural scenes, when the number of stimuli observed by the subject is significantly smaller (∼100). Using simulations, we show that our method can recover the structure and spatial location of noisy, non-linear receptive fields from the recorded saccade point locations on observed natural stimuli together with the related neural activity. Our method relies on the use of a statistical test that can reliably predict whether the estimated receptive field is significant or not. Finally, we present receptive fields extracted using our method from real physiological recordings. doi:10.1016/j.neures.2009.09.490
P1-h09 Visual experience modifies the neural dynamics for orientation processing in cat visual cortex Kazunori O’Hashi 1,2 , Amiram Grinvald 1 , Shigeru Tanaka 2 1
Weizmann Institute, Israel;
2
RIKEN BSI, Japan
Orientation selectivity of visual cortex is highly adaptive to visual experience in early postnatal life. Orientation-restricted environment during the critical period can transform the orientation representation to respond to the exposed orientation preferably. It remains unclear, however, how orientation signals are processed by the altered visual cortical circuits. In this study, we examined spatiotemporal dynamics
Abstracts for orientation processing in the visual cortex of the goggle-reared animals using voltage-sensitive dye imaging. As a result, we found that the response onset to the exposed orientation was faster than the others, whereas the responses to unexposed orientations showed brief but strong hyperpolarization followed by depolarization. In addition, orientation preferences were temporally unstable at late period of visual stimulation. After 1-week normal viewing, however, the temporal dynamics of visual responses were restored to those in normally reared animals. These findings suggest that the visual cortical circuits change the way of information processing depending on visual experience. doi:10.1016/j.neures.2009.09.491
P1-h10 Neural correlates of intersubject variability of continuous flash suppression Hiroyuki Yamashiro 1,2 , Hiroki Yamamoto 1 , Jun Saiki 1 , Hiroaki Mano 1,4 , Masahiro Umeda 3 , Chuzo Tanaka 4 1
Grad. Schl. of Human & Environmental Stud., Kyoto Univ, Kyoto, Japan; JSPS, Tokyo, Japan; 3 Department of Med. Informatics, Meiji University of Integrative Med., Kyoto, Japan; 4 Department of Neurosurgery, Meiji University of Integrative Med., Kyoto, Japan
2
In continuous flash suppression (CFS), a monocular target stimulus can be suppressed from visual awareness for up to several minutes by a presence of a highly dynamic mask in the other eye. Interestingly, the durations of suppression depend greatly on subjects, which range from several seconds to minutes. To explore the neural correlates of the intersubject variability of CFS, we measured brain activity using fMRI, while subjects viewed a rotating target rendered invisible by CFS. We found a strong correlation between the effects of CFS and the retinotopic responses to the invisible target in extrastriate visual areas (V3, V3A, V4, V7, MT, LOc): the less susceptible subjects were to CFS, the larger responses were evoked. This suggests that extrastriate visual areas play a critical role in generating and suppressing visual awareness. doi:10.1016/j.neures.2009.09.492
P1-h11 Monkeys perceive orientations relative to vertical Masumi Wakita Primate Res Inst, Kyoto Univ, Aichi, Japan Monkeys readily discriminate cardinal orientations (horizontal and vertical) from oblique orientations. In addition, they recognize cardinals without confusing with any other orientations. Intuitively, cardinals may be a reference to judge degree or direction of obliques. However, functional difference between horizontal and vertical remains unclear. Here, five monkeys underwent two discrimination training between 22.5◦ and 157.5◦ gratings (horizontal condition; stimulus separation was 45◦ across horizontal axis), and between 67.5◦ and 112.5◦ gratings (vertical condition; stimulus separation was 45◦ across vertical axis). After each training session, their performance was tested using gratings from 0◦ to 157.5◦ with a step of 22.5◦ . Consequently, generalization tests revealed that animals showed wider generalization curve in the vertical condition compared with the horizontal condition. In other words, monkeys perceived a distance between 67.5◦ and 112.5◦ was smaller than that between 22.5◦ and 157.5◦ . Thus, they perceive orientations relative to vertical axis. doi:10.1016/j.neures.2009.09.493
P1-h12 Higher-order structures in natural scenes Takuma Tanaka 1 , Toshio Aoyagi 2 , Takeshi Kaneko 1 1 2
Department of Morphogical Brain Science, Kyoto Univ, Kyoto, Japan; Graduate School of Informatics, Kyoto University, Kyoto, Japan
It has been shown that the selectivity of neurons in lateral geniculate nucleus and simple cells in the primary visual cortex can be reproduced by the models which optimize the information transmission between the input and output units. Recently, Karklin and Lewicki (2009) reported that complex cell-like selectivity can be selforganized through the learning of the probability distribution of natural images. We present a model in which the activity of neurons represents the probability of the present input. Using the output of simple cell-like linear filters as the input, we found that neurons in this model after learning exhibited the orientation preference and translation invariance, both of which are properties of complex cells. We conclude that complex cells detect and represent the higher-order structures of the probability distribution of natural scenes. doi:10.1016/j.neures.2009.09.494
S109
P1-h13 Transfer of synesthetic experience Aleksandra Mroczko 1,2 , Thomas Metzinger 2,3 , Wolf Singer 1,3 , Danko Nikolic 1,3 1
Max Planck Institute for Brain Research, Germany; 2 University of Mainz, Germany; 3 Frankfurt Institute for Advanced Studies, Germany
In synesthesia, a certain stimulus (e.g. grapheme) is associated automatically and consistently with a stable perceptual-like experience (e.g. color). We found that grapheme-color associations can transfer to novel graphemes after only a 10-minute writing exercise. Most subjects experienced synesthetic associations immediately after learning a new Glagolitic grapheme. Using a Stroop task, we provide objective evidence for the creation of novel associations between the newly learned graphemes and synesthetic colors. Also, these associations generalized to graphemes handwritten by another person. The fast learning process and the generalization suggest that synesthesis begins at the semantic level of representation with the activation of a certain concept (the inducer), which then, uniquely for the synesthetes, activates representations at the perceptual level (the concurrent). doi:10.1016/j.neures.2009.09.495
P1-h14 Target-specific connectivity enhances the stability of activity dynamics in a layered cortical network model Tobias C. Potjans 1,2 , Markus Diesmann 2,3,4 1
2 Research Center Juelich, Germany; RIKEN Computational Science Research Program, Wako-shi, Japan; 3 RIKEN Brain Science Institute, Wakoshi, Japan; 4 Bernstein Center for Computational Neuroscience, Freiburg, Germany
The local cortical network consists of distinctively interconnected layers. In addition to the well-established feedforward pattern of connections (L4 to L2/3 to L5 to L6), specific feedback connections have been identified that specifically select interneuronal targets. We employ large-scale simulations to investigate the dynamical consequences of these target specific connections. Our model consists of 80,000 neurons and reflects the layer-specific connectivity established in anatomical and physiological studies. We find that target-specific connections enhance the stability of network activity. Furthermore, target specificity increases the sensitivity for time-dependent signaling: only these networks exhibit strong, temporally locked responses to transient thalamic inputs. doi:10.1016/j.neures.2009.09.496
P1-h15 A gaze prediction model considering the inhibition of return based on the saliency map theory Taiki Gochi, Takeshi Kohama Grad. Sch. Biology-Oriented Science and Technology, University of Kinki, Wakayama Japan Previous saliency-based visual attention models deal with the mechanism of attentional shifts, and such models can predict the distribution of gaze movements based on the saliency map. Nevertheless, repeating a constant movement pattern in the prediction model poses a problem. Therefore, in order to improve this unrealistic prediction, it is necessary to consider the abstraction of the inhibition of return (IOR). IOR refers to the suppression of the processing of objects that have recently been gazed at attentively and prevents the viewer from watching the same objects repeatedly. In previous models, IOR operates at constant intensity in the higher stage of processing; however, the intensity of inhibition must depend on the importance of the gazed object. In this study, we modified the IOR operation by the Gaussian proportion decrement with random peak values in the lower-stage processing. As a result, the predictions of gaze distribution were more sparse and realistic. This indicates that the inhibition of return might be a low-order operation at variant intensity. doi:10.1016/j.neures.2009.09.497
P1-h16 The influence of figure components and their motion direction on motion position perception Azusa Deguchi, Akitoshi Hanazawa Department Brain Sci. Eng., Kyusyu Institute of Technology, Kitakyusyu, Japan We found a new optical illusion. We made a figure consisting of one long line and many short lines, which are perpendicular to it, and turned this figure around the middle point of the long line. Then rotating position of the short lines was perceived behind the long line. This phenomenon can be thought as follows. For the short line, because motion direction is parallel to the axis of the line, motion is detected only at the end points. On the other hand, for the long line, because motion direction is