Modelling Fly Motion Vision - Computational Neuroscience: A Comprehensive Approach - page 413

network from the sensory input when rotational flow-fields stimulate the eyes, but
not when translational stimuli occur. The flow-field selectivity of LPTCs therefore
seems not only determined by their feed-forward connectivity, but also by the intrin-
sic wiring within the network formed by LPTCs in both hemispheres. Of course,
the circuit diagram presented in Figure 14.1 only contains a small number of the
existing lobula plate neurons (5 out 60 on each side!) where such connections have
been studied in detail to date. In particular, no neurons sensitive to vertical image
motion are included there. More investigations concerning the connectivity amongst
such VS-cells and between the horizontal and vertical cells are presently under way
(Haag and Borst, in preparation) possibly providing an explanation for the complex
flow-fields measured in many neurons of the lobula plate [63, 64].
To gain a further understanding in the functional consequences of such network
interactions, modelling work was started. As a first step, a network model of the lob-
ula plate was built using single-compartment models of each cell in the circuit. After
adjusting the connectivities to the experimental data, the individual circuit elements
revealed similar responses to binocular motion stimuli as their natural counterparts
( Figures 14.13 and 14.14) . First of all, in response to rotational motion stimuli, all
neurons modulate their response strongly when the stimulus is switched from clock-
wise to counter clockwise rotation (Figure 14.13). In this case, the internal connec-
tions of the circuit amplify the excitation levels imposed onto the neurons by the
local motion detectors, i.e., when a neuron on one side is excited, its counterpart on
the other side becomes inhibited. The situation changes when instead of rotational
motion translational stimuli are presented (Figure 14.14): now, the modulation is
weaker in HS-cells and CH-cells when the stimulus switches from contraction to ex-
pansion. In particular in CH-cells, the membrane potential departs only little from
resting during stimulation either way. In this case, the internal connections work
against the feed forward signals coming from the local motion detectors and, thus,
reduce the responses substantially. While these simulations represent only a starting
point, future studies will investigate how these internal connectivities affect the re-
sponse behavior of the circuit elements once more critical stimuli are presented e.g.,
low light levels or low contrast patterns where noise becomes an issue.
14.3.4 Dendro-dendritic interactions
In contrast to the lobula plate neurons covered so far which all respond strongest
to large-field motion, another group of tangential cells has been described which
respond best to small moving objects or relative motion (called FD- or CI-neurons;
[25, 38]. These cells receive inhibitory input when contralateral back-to-front motion
is additionally displayed to an excitatory ipsilateral front-to-back stimulus. There is
evidence that the vCH-cell is responsible for conveying this type of inhibitory input
since the FD1-cell was shown to loose its inhibitory input after photo-inactivation
of the vCH-cell [95]. To what extend the CI-cells are also inhibited by the vCH-
cell and whether the inhibition of the FD1- and/or CI-neurons by the vCH-cell is
also responsible for their ipsilateral small-field tuning, i.e., their preference for small
objects moving, and thus represent the 'pool cell' postulated in the original models