Despite the fact that several highly successful computational models for motion detection have been established in flies and extended to applications in other fields like computer vision, several key processing steps in the optic lobes of the fly are still unresolved, especially in regard to their neuronal implementation. This is partly due to the fact that these computations take place at the dendrites of cells or in densely packed neuropils that are difficult to access electrophysiologically.

I tried to shed light on these intermediate neuronal computations by examining how visual features like velocity, direction and orientation are represented at different stages of motion vision. For this, I used calcium imaging to label different classes of neurons and quantify their activity in vivo while the animal was stimulated with visual motion patterns.

The first step of the study looks at dendritic integration of motion signals by the large lobula plate tangential cells. These neurons integrate and process inputs from large parts of the visual field in a fashion that is characteristic for each of these in¬dividually identifiable cells, with specific preferences for global motion directions. I was able to show that the dendrites of this cells display a patterns of localized motion preferences that forms a dendritic map of the visual field, thus creating a selective filter for complex motion patterns. Also, I could show that distributed inputs from contralateral neurons act as additional influences on these retinotopic patterns, shaping their layout in a way that could not be predicted through axonal recordings alone.

The second part of the study focuses on the putative input elements of the tangential cells, the cells of the medulla. These neurons form a complex retinotopic mosaic of columns which are connected by lateral interactions, and are believed to be one of the key neuropils for the extraction of image features in the fly visual system. Through a novel population staining method, I was able to introduce a calcium in¬dicator into these cells and examine their responses to motion as well as various flicker stimuli. The stained neurons, which consisted of columnar elements as well as tangential processes, exhibited strong motion responses, but did not show a distinct preference for a single motion direction like the cells of the lobula plate. Instead, many of them showed symmetry in their motion responses, preferring one pattern orientation over the other but not differentiating between opposite motion directions. Also, when stimulated with either bright ("on-") or dark ("off-") edges, the neurons responded to both, with spatial composition of inputs depending on the dye injection site: Some cells simply integrated spatially on- and off- signals with similar receptive fields, while other populations displayed an integration over spatially separated channels.

In the third chapter, I review some of the techniques developed and used during the study in the context of other methods for examining and influencing cellular acti¬vity through the application of calcium indicators, caged calcium and phototoxic dyes, which allow monitoring and manipulation of cellular networks on a single cell basis.