Neural encoding of behaviourally relevant visual-motion information in the fly

Abstract

The goal of neuroethology is to explain behaviour in terms of the activity of nerve cells and their interactions. This can only be achieved if the experimental animal can be analysed at different levels ranging from behaviour to individual neurons. Cellular mechanisms underlying processing of neuronal information are frequently analysed using in vitro preparations where artificial stimulation replaces the natural sensory input. Although such studies provide fascinating insights into the complex computational abilities of neurons [1], the results may not be extrapolated easily to in vivo conditions, where the range of response amplitudes of neurons and their temporal activity patterns may differ considerably from artificially induced activity. In systems such as the retina of the horseshoe crab [5,6], it is now feasible to analyse the neuronal representation of visual input as it is experienced during behaviour (reviewed in Refs [7,8]). Until now, however, in most systems the underlying neuronal mechanisms have been difficult to unravel.In the fly it is possible to employ both quantitative behavioural approaches as well as in vivo electrophysiological and imaging methods to analyse how behaviourally relevant visual input is processed [9][10][11][12][13][14][15][16][17][18][19][20]. Although the latter techniques are mainly employed in the blowfly, which is relatively big, they are complemented by studies of the smaller fruitfly, Drosophila, where a broad range of genetic approaches can be applied to dissect the visual system in an increasingly specific way [21,22].We review recent progress on the encoding of optic-flow information in the blowfly. Optic flow is an important source of information about self-motion and the three-dimensional layout of the environment, not only for flies but for most moving animals including humans (Box 1, [4,[23][24][25]). Flies exploit optic flow to guide their locomotion [13] and to control compensatory head movements [26], and understanding the computational principles underlying optic-flow processing in flies could provide insights into visual-motion analysis in general.Information processing in visual systems is constrained by the spatial and temporal characteristics of the sensory input and by the biophysical properties of the neuronal circuits. Hence, to understand how visual systems encode behaviourally relevant information, we need to know about both the computational capabilities of the nervous system and the natural conditions under which animals normally operate. By combining behavioural, neurophysiological and computational approaches, it is now possible in the fly to assess adaptations that process visual-motion information under the constraints of its natural input. It is concluded that neuronal operating ranges and coding strategies appear to be closely matched to the inputs the animal encounters under behaviourally relevant conditions.

Global versus local adaptation in fly motion-sensitive neurons

“…This aftereffect, the waterfall illusion, is not unique to humans: elegant behavioural work has shown that it is also experienced by flies (Götz & Wenking 1973;Srinivasan & Dvorak 1979), indicating that motion adaptation is also relevant to these animals (Clifford & Langley 1996). In flies, the physiological counterpart of this phenomenon can be found in motion-sensitive neurons in the third visual neuropile (lobula plate), most of which sum motion signals across large areas of visual space (Hausen 1984;Krapp & Hengstenberg 1996;Egelhaaf et al 2002). Similarly to macaque MT neurons (Kohn & Movshon 2003, these cells adapt their response characteristics to prolonged stimulation (Srinivasan & Dvorak 1979;Maddess & Laughlin 1985;de Ruyter van Steveninck et al 1986;Fairhall et al 2001).…”

Directional Selectivity in the Simple Eye of an Insect

“…We observed the same inseparabilities in the UV STRFs of L-neurons. This type of directional computation is distinct from that seen in fully opponent directionally selective cells, like lobula plate tangential cells in flies (Hausen, 1982a,b;Single et al, 1997;Borst and Haag, 2002;Egelhaaf et al, 2002) or complex cells in the mammalian visual cortex (Emerson et al, 1992) that encode directionality in the mean of the response.…”

Local visual homing by matched-filter descent in image distances

“…A matched filter is a concept in neuroethology and describes the spatial layout of some population of receptors that is matched to a certain aspect of the task (Wehner 1987). For example, preferred directions for arrays of elementary motion detectors seem to corre- spond to flow fields induced by particular movements of flies, e.g., by roll movements (Egelhaaf et al 2002;Krapp 2000;Franz and Krapp 2000;Franz et al 2004). Our matched filters have the typical structure of flow fields for pure translations in the horizontal plane: a focus of expansion, a focus of contraction, and nearly horizontal flow between them (Fig.…”