SYMPOSIA...
|
| The role of feedback in visual perception |
| Mark Williams - Macquarie University, Australia
Co-authors: Chris Baker, Hans Op de Beeck, Sabin Dang, Christina Triantafyllou, Nancy Kanwisher
Location-Invariant Object Information in Foveal Retinotopic Cortex
According to standard theories of object perception, information in early stages of the visual hierarchy is tied to stimulus location, with more abstract ‘position-invariant’ representations of object shape achieved only at later stages. Here we present a new phenomenon that challenges this standard view: the pattern of fMRI response in foveal retinotopic cortex contains position-invariant information about objects presented outside the fovea. Subjects fixated centrally while viewing three categories of novel objects. In each trial, two objects were presented simultaneously in diagonally opposite peripheral retinal locations. These two objects were always from the same object category, and subjects were asked whether the two objects were identical or subtly different exemplars of that category. We then used multivariate pattern analysis methods to ask whether information about object category was present in the pattern of response across voxels in each of several cortical regions of interest (ROIs). Our experiments show that the pattern of fMRI response in foveal retinotopic cortex contains information about the category of objects presented outside the fovea, in the visual periphery. The pattern in foveal cortex is similar regardless of whether the stimuli are presented in the same or different peripheral locations, thus it is largely position-invariant. This object information in foveal retinotopic cortex is behaviorally relevant: i) it is present during a discrimination task on objects presented in the periphery, but not a color discrimination task performed on the same stimuli, and ii) stronger information in foveal cortex is correlated across subjects with higher task performance Control experiments ruled out differential eye movements across object categories, activation from the fixation cross itself, or spillover activation from peripheral retinotopic cortex or from LOC. Instead, our data strongly suggest that position-invariant object information from higher cortical areas is fed back to foveal retinotopic cortex, improving task performance.
|
|
Christian Ruff - University College London, UK
Top-down control of visual processing: The role of anatomical feedback connections
It is increasingly recognized that processing within occipital visual cortex does not only reflect retinal input, but may be influenced by areas outside the conventional visual system. For instance, it is often proposed that a putative fronto-parietal ‘attention network’ may modulate visual processing via back-projection influences. However,
neuroimaging experiments in humans have so far only provided indirect correlational evidence for such functional interactions, and have not revealed distinct contributions of different frontal or parietal sites. I will present a series of experiments that combined transcranial magnetic stimulation (TMS) of human frontal eye-field (FEF) or intraparietal sulcus (IPS) with concurrent functional magnetic resonance imaging (fMRI), to show directly that feedback connections from these fronto-parietal regions have the capacity to elicit systematic and specific activity changes in human retinotopic visual cortex. The FEF- or IPS-TMS influences on visual processing differed in several respects (such as their precise spatial topography across retinotopic visual field representations or their dependence on concurrent visual input), underlining the specificity of functional influences from the stimulated sites on activity in visual cortex. Moreover, all effects were more marked during right-hemisphere (vs left-hemisphere) stimulation, mirroring the hemispheric asymmetry observed for the effects of lesions or TMS to such sites on perfromance in visual tasks. Finally, the functional relevance of these occipital activation changes for visual perception was also demonstrated in a separate psychophysical experiment, where we found that FEF TMS elicited changes of contrast sensitivity in different parts of the visual field that directly mirrored the pattern of occipital activity changes observed during fMRI of FEF TMS. In sum, our data demonstrate directly that frontal and parietal regions of the human brain can have causal, specific, and functionally relevant influences upon activity in human retinotopic visual cortex, which may provide a neural substrate for top-down influences on visual processing.
. |
|
Fang Fang - Peking University, China
Attention-dependent contextual effects in human early visual cortex
Contextual effects are ubiquitous in vision and reveal fundamental principles of visual information processing. Here, I present two fMRI studies investigating contextual effects in human early visual cortex. The first study revealed that perceived image size can modulate retinotopy in human primary visual cortex (V1). We presented two tori with the same physical size at close and far apparent depths in a three-dimensional scene of a hallway and walls. When fixating its center, the far torus appeared to be larger and occupy a more eccentric portion of the visual field, relative to the close torus. Using high-resolution fMRI, we found that the retinotopic distribution of V1 activity induced by the far torus was also shifted towards a more eccentric representation of the visual field, while that induced by the close torus was shifted towards the foveal representation, consistent with their perceptual appearances. In the second study, we used fMRI adaptation to investigate sensitivity to border ownership in human early visual cortex. Two stimuli were generated by modifying a black/white square-wave radial grating. In one stimulus, the white stripes were slightly longer than the black stripes in the radial direction. This provided contextual information that caused the borders between the white and the black stripes to appear to belong to the white stripes. In the other stimulus, the black stripes were slightly longer than the white stripes, which caused the borders to appear to belong to the black stripes. Subjects adapted to one of these two stimuli and were then tested with both stimuli. Regions of interest in the early visual cortex were confined to areas corresponding to an annular interior part of the radial grating. This part was locally identical across the two stimuli, but as a consequence of the difference in the contextual information, the borders between the white and the black stripes were perceived to belong to either the white or the black stripes. We found a strong adaptation effect in V2, but a weak effect in V1, which suggest that V2 is a critical area for the processing of border ownership. In both studies, the contextual effects occurred only when subjects directed their attention to the stimuli, which imply that contextual information processing depends on feedback from higher-level visual areas per se.
|
|
Toshiyuki Hirabayashi - The University of Tokyo, Japan
Co-authors: Daigo Takeuchi, Hironori Kasahara, Keita Tamura, Yasushi Miyashita
Cross-correlation and spectral analyses of spike trains in macaque inferotemporal cortex for different parts configurations within a whole object
Neurons in the inferior temporal (IT) cortex have been shown to change their response selectivity by repetitive learning of complex shapes. However, how the IT neurons represent novel shapes that were composed of combination of familiar shapes has not been elucidated. In the present study, we prepared 40 X 40 x 40 facial-feature-like parts (eye-, nose- and mouth-like parts) to construct 64,000 face-like objects (FOs) and non-face-like objects (NFOs) in which the same set of parts were randomly arranged. We conducted multiple single-unit recordings from the macaque IT cortex to examine spike correlation in response to FOs and NFOs composed of the same parts. FOs and NFOs were optimized to elicit as much responses as possible from both the neurons of the recorded pairs by choosing adequate combinations of parts. Of the 134 neurons showing significant responses to both their optimal FOs and NFOs, 30 pairs of neurons revealed significant spike correlation in response to either their optimal FOs or NFOs. We found that spike correlation was significantly stronger for FOs than for the corresponding NFOs (P < 0.01). Firing rates of these neurons for their optimal FOs and NFOs were not significantly different (P > 0.9). These results suggest that IT neurons exhibit dynamic modulation of spike correlation depending on the spatial configuration of parts within a whole object.
Neuronal interactions have often been reported to exhibit some spectral characteristics and they might play an important role in visual information processing. We thus further conducted spectral analysis for the spike trains of the same pairs of IT neurons. We found that these neuron pairs also revealed FO-dominant spike-spike coherence in the gamma-range (40-100 Hz, P < 0.02). Moreover, we divided the neuron pairs into two groups on the basis of the characteristics of their functional connectivity: neuron pairs with center peak or displaced peak on the cross-correlogram. Phase analysis of spike coherence showed that neuron pairs with a displaced peak revealed a larger phase of coherence than neuron pairs with a center peak (P < 0.05). These results suggest that IT neurons show dynamic modulation of spike coherence in the gamma range with various phase lags depending on the property of functional connectivity.
Altogether, IT neurons revealed dynamic modulation in their mutual interactions depending on the presented stimulus, and these interactions were detectable in the analysis of either the time or frequency domain of spike trains. |
| |
