LIN: Forschungsabteilungen > Akkustik, Lernen, Sprache > Unterpunkt Ebene 3 > Unterpunkt Ebene 4
Titel: LIN Layout
Neural mechanisms of visual search
Consider the picture showing newspaper vending machines somewhere in San Francisco . Say you want to find the yellow machine with the red base (blue circle). How does the visual system solve this task? First, beause you only know that you must search for a target that is yellow and red, the visual system will need to start from that color information to locate the item. Here it already encounters a problem. Obviously, there are other colors such as blue and green which are also detected by the visual system. How is it that yellow and red, but not blue and green, are made to guide your search? One major focus of our group's research is devoted to the investigation of the brain mechanisms that solve this and related problems. See Hopf et al. (2004a) for first results.
Assume the visual system has now solved the above problem of selecting relevant colors and comes up with a representation such as the one in the figure below. Object locations with blue or green vending machines are ‘grayed out', and do not interfere with the search process anymore. However, there is another problem. Yellow and red appear in more than one vending machine. How does the visual system pick the one machine that is both, red and yellow, without also selecting the other red or yellow machines. In more technical terms, the attempt to identify the target's location based on the spatial distribution of relevant colors encounters ambiguities in location information. How does the brain solve such ambiguities?
One important source of such ambiguities is that receptive fields (RF) of neurons become larger and larger at higher levels of the visual processing hierarchy - a consequence of their considerable upstream convergence. Higher-level cells, therefore, ‘see' larger parts of a scene than lower level cells. Accordingly, more and more objects appear within one RF at higher processing levels, but the cell cannot distinguish between these objects. Now, coming back to our newspaper vending machines, the yellow circle illustrates a possible RF of such a higher-level cell. While the lower-level cell (small blue RF) will correctly identify the target, the higher-level cell will also assume to have found the target, as it doesn't know whether yellow and red are part of one or different objects*.
Research of our group (Hopf et al.,2000, 2002a) provides evidence that the suppression of unwanted neural representations is one important mechanisms that serves to resolve such ambiguities.
* This example serves to illustrate the problem of ambiguous coding. We do not assume that there are color responsive cells tuned for a combination of red and yellow.
Bottom-up perceptual grouping and top-down attentional orienting
It is a common experience that one is able to intentionally direct attention to a selected part of a visual scene (top-down attentional control). However, perceptually salient stimuli can also attract attention in a bottom-up manner. With this project we address the question how top-down and bottom-up processes interact in the visual system to guide focusing. See Hopf et al. (2004b) for first results.
Recurrent processing in attentional orienting
In this new project we investigate the role of feed-back processing between higher- and lower-level visual areas (e.g. V1) in attentional orienting. A ttention modulates neural activity in early visual areas such as the striate cortex (V1). ERP/MEG recordings, however, revealed that these modulations appear delayed and generally not before 150ms after stimulus onset. This indicates that attention doesn't impact the initial feed-forward sweep of processing in V1 and suggests a feed-back origin of it's modulatory effects. What remains to be clarified is the particular processing role of these recurrent activations.