Wicklein, Martina
Coauthors(s): TJ Sejnowski CNL Salk Institute, La Jolla CA 92093
The Salk Institute
CNL
10010 North Torrey Pines Road La Jolla, California 92037-1099 USA
www.cnl.salk.edu/~martina/

PERCEPTION OF LOOMING IN THE HUMMINGBIRD HAWKMOTH MANDUCA SEXTA (SPHINGIDAE, LEPIDOPTERA)

When foraging hawkmoths hover in front of flowers during feeding they use visual depth cues to control and stabilize their distance to the flowers. The moths are able to maintain their position even while the flower moves, irrespective of the direction of movement. Visual perception of depth change can reliably be mediated monocularly by looming, the apparent size increase of an approaching object. Looming sensitive neurons are found throughout the animal kingdom including sphingids (Wicklein and Strausfeld in press), grasshoppers (Rind, 1990, Gabbiani et al. 1999), pigeons (Sun and Frost, 1992), and monkeys (Graziano et al., 1994). Intracellular recordings In M. sexta have identified two classes of cells that respond to looming stimuli and report both the approach and retreat of an object. The cells (class 1 and 2 neurons) compute looming in two fundamentally different ways. Class 1 neurons respond only to an approaching or retreating disc but not to rotating spirals or moving gratings. We conclude that they measure the change of perimeter/edge length of the object. Class 2 neurons respond to approaching or retreating disc, inward or outward rotating spirals and moving gratings. This suggests that class 2 cells measure the expansion/contraction flowfield to determine the approach or retreat of an object. Class 1 neurons have cell bodies in the optic lobes and wide dendritic arborizations in the innermost stratum of the medulla, within the outer stratum of the lobula, and throughout all layers of the lobula plate. Their terminals invade the ipsilateral optic foci of the posterior slope. The large cell bodies of class 2 neurons are situated posteriorly between the calyces of the mushroom bodies. Their dendritic trees reside in the posterior slope neuropil on one or both sides of the brain; the axons extend out to the ipsi- or contralateral medulla with wide terminals in the outermost medulla stratum. We created a network model to help us understand the underlying computational principals leading to looming detection. It is based on the hypothesis that class 1 cells infer change of depth by computing change of perimeter length and incorporates the anatomical and physiological properties of this cell class. To assess the validity of the model we test model and neuron with the same stimuli. Comparing the model output with recordings from class 1 neurons, we conclude that the model captures many of the essential properties of the physiological data gathered from class 1 cells. Both the cells and the model show sustained responses as long as the looming/receding stimulus is presented. Neither the cells nor the model respond to moving gratings and the response to expanding and contracting objects is qualitatively similar. Literature Gabbiani F, Krapp HG, Laurent G (1999) J Neuroscience 19(3):1122-1141 Rind FC (1990) J Exp Biol 149:21-43. Sun H, Frost BJ (1998) Nature Neuroscience 1 (4):296-303 Wicklein and Strausfeld (in press in J Comp Neurol)