|
|
 |
|
| Matthew Self |
 |
Anatomy department, University College London, Gower Street, London, WC1E6BT
e-mail: m.self@ucl.ac.uk |
Poster Presentation: |
|
| THE POSTERIOR REGION OF THE LATERAL OCCIPITAL COMPLEX INTEGRATES COLOUR AND MOTION INFORMATION TO CONSTRUCT SIMPLE SHAPES.. |
| M.W.Self and Semir Zeki | |
|
Wellcome departemt of imaging neuroscience, London, UK. |
|
| Colour and motion are processed by separate neural systems in the brain. Using psychophysics and fMRI, we investigated how the distributed information from these functionally specialized systems is integrated to allow the recognition of objects defined by both colour and motion. In the psychophysical experiment we presented to 3 subjects shapes composed of thousands of coloured kinetic dots, and controlled their visibility by varying the coherence of either colour or motion or both. Our results show that if the same shape is defined by both the coloured and the moving dots (transmodal shapes) they can be recognised at lower coherence values than shapes composed from either colour or motion alone (unimodal shapes). This demonstrates that shapes can be rendered suprathreshold by the combination of subthreshold stimuli from different modalities. Croner et al have shown that the visual system is able to segment an image on the basis of uniform colour to improve the detection of a motion signal. It is possible that segmenting out the signal dots on the basis of uniform colour in the transmodal condition improves the detection of the shape edges. To test this hypothesis we designed a transmodal stimulus in which the colour and motion information was non-overlapping in space (i.e. at 50% coherence half the dots move in a uniform direction of motion and the other half have a uniform colour). Performance on this task was identical to the normal, overlapping transmodal stimulus. This indicates that the improvement in performance in the transmodal task is not due to image segmentation but must be due to a modality independent integrating neural process with a relatively coarse spatial resolution. This result therefore demonstrates that the brain does not necessarily use the most salient modality but rather integrates both modalities together to construct shapes. We used event-related fMRI and the same stimuli to identify whether this is done in specific areas, and found that the only area of the brain which was more strongly activated (p < 0.05, corrected) by transmodal shapes than unimodal ones (at the same psychophysical performance level) was the posterior part of the lateral occipital complex (LO). These results suggest a simple model for shape recognition whereby edge information is processed by functionally specialised, modality specific areas (i.e. V5, V3 complex for motion borders, V4, V2 for colour) after which these edges are integrated in a cue-invariant area (LO) to construct shapes. In this way LO is able to recognise transmodal shapes that would remain invisible to a unimodal system.
|
|
|
 |
|
|
|
