Abstract
When the gaze is moved quickly, vision becomes blurred and the way in which three-dimensional structures are seen can be affected1. How can these effects be minimized? During flight, blowflies (Calliphora vicina) turn both the head and the thorax very quickly, producing gaze shifts which affect vision. Here we show that blowflies reduce the effects of thorax movements on vision by moving the head later, and more quickly, than the thorax. This reduces the time in which gaze is shifting relative to the surroundings, and maximizes the time available for analysing the surroundings.
Main
Gaze shifts present a challenge to vision for at least two reasons. First, it takes time for optic signals to be integrated by the photoreceptors in the eye. If the gaze is shifted, this integration time will result in a blurred image being produced, particularly during fast gaze shifts. Second, the resulting rotational flow of optic signals will interfere with the pattern of optic flow that normally reveals the three-dimensional structure of the surroundings during linear motion1.
Blowflies can move their two compound eyes by moving the head2,3. The head, which is connected to the thorax by a neck, has appreciable freedom of movement under muscular control4,5. We measured the position and orientation of thorax and head during (almost) free flight by using a modified search coil technique6. The thorax usually changes course through a series of short, fast turns7, with gradual course changes occurring less frequently. The head usually turns in synchrony with the thorax but at a higher angular speed (Fig. 1a), with a pattern similar to that of fast eye movements in vertebrates8.
The average angles of yaws and rolls (Fig. 1b) during head-thorax turns of 10-20 degrees to the left (the most common yaw angle; turns of other angles follow similar time courses) are shown in Fig. 1c, d. A turn starts with a rotation of the thorax. The thorax yaw (Fig. 1c) is not immediately accompanied by a head yaw relative to the surroundings, as it is first compensated by a rotation of the head in the opposite direction relative to the thorax. After about 10 milliseconds, the head starts to rotate in the same direction as the thorax. It reaches its highest yaw velocity relative to the thorax at about the same time as the thorax reaches its highest yaw velocity relative to the surroundings. When the head approaches its final orientation relative to the surroundings, it decelerates, and finally rotates in the opposite direction to the thorax. The result is a stabilized yaw of the head roughly 10 milliseconds before the thorax rotation is finished. As a result of the shortening of the head turn compared with the thorax turn, the period of blurred vision during course changes is reduced by roughly one-third.
Like aeroplanes, flies usually make banked turns7, in which large changes in yaw are accompanied by large rolls of the thorax. The head can partly correct the visual consequences of this by performing a counter roll9 (Fig. 1d). The residual roll movement of the head relative to the surroundings is small (typically less than 5 degrees). Rotations in the pitch direction (not shown) are also usually small.
We can deduce the visual consequences of coupled thorax and head turns from Fig. 2. We divided the total flight time into episodes during turns and episodes between turns. During turns, the yaw velocity of the head is higher than that of the thorax (Fig. 2a). As the photoreceptor integration time for the light level during the measurements was roughly 10 milliseconds, and the angular resolution of the photo-receptors is 1-2 degrees, angular velocities higher than 100-200 degrees per second will cause significant blur in the fly's visual system. The high angular velocity of the head during turns therefore makes it impossible to see details. Between turns, however, the head is more stable than the thorax (Fig. 2b), mostly within a velocity range that gives little blur attributable to rotation.
The triphasic movement of the head in the yaw direction (Fig. 1c) probably acts to maximize the periods of stable gaze by minimizing the duration of the gaze shift. The extra angular velocity produced by the neck muscles is effective because it peaks at about the same time as the peak of the thorax angular velocity. The effective duration of the most common gaze shifts is not much longer than the integration time of the photoreceptor. Because at least one integration period is lost by making a gaze shift anyway, the roughly 1.5 integration periods lost during a typical blowfly turn seem to be a reasonable compromise between minimal visual impediment and very fast movement.
References
Koenderink, J. J. Vision Res. 26, 161–180 (1986).
Land, M. F. Nature 234, 299–300 (1973).
Geiger, G. & Poggio, T. Biol. Cybernet. 25, 177–180 (1977).
Hengstenberg, R. in The Head-Neck Sensory Motor System (eds Berthoz, A., Graf, W. & Vidal, P. P.) 49-55 (Oxford Univ. Press, 1992).
Strausfeld, N. J., Seyan, H. S. & Milde, J. J. J. Comp. Physiol. A 160, 205–224 (1987).
Schilstra, C. & van Hateren, J. H. J. Neurosci. Methods (in the press).
Wagner, H. Phil. Trans. R. Soc. Lond. B 312, 527–551 (1986).
Carpenter, R. H. S. Movements of the Eyes (Pion, London, 1988).
Hengstenberg, R. J. Comp. Physiol. A 163, 151–165 (1988).
Haslwanter, T. Vision Res. 35, 1727–1739 (1995).
Author information
Authors and Affiliations
Supplementary Information
Methods
Figure 1. a, The position and orientation of both thorax and head were measured simultaneously by a double version of the modified search coil technique described in ref (6). Briefly, 3 orthogonal pairs of magnetic field coils (driven at 50, 68 and 86 kHz, two fields homogeneous, one gradient) surrounded a cage of 40x40x40 cm3; in this cage blowflies flew with an average speed of 0.58 m/s. On the fly's head and thorax, two systems of 3 orthogonal sensor coils were mounted (weights 0.8 and 1.6 mg; cf. 8 mg for a typical blowfly head). The voltages induced in these coils were measured with two sets of 9 lock-in amplifiers. The resulting system yields position and orientation with a temporal resolution of 1 ms, a spatial resolution of approximately 1 mm, and an angular resolution of approximately 0.5 deg. The head coils were connected to the thorax via a flexible wire loop; control experiments on tethered flies, performed by varying the mechanical load of coils and loop, showed negligible influence of coils and loop on head rotations. The head and thorax coils were connected to the bottom of the cage via a thin cable (12 leads of 12 micrometer copper wire, total weight 10 mg; cf. 80 mg for a typical blowfly). Control experiments performed by varying the weight of the cable showed no interference with normal flight. The walls of the cage were covered with patterns of branches and leaves; average luminance was 150 cd/m2 for the walls, and 800 cd/m2 for the ceiling. b, Rotations are described by an ordered sequence of yaw, pitch, and roll (Fick coordinate system10). In order to comply with aeronautical sign conventions, we define yaw angle=−θF, pitch angle=−φF, and roll angle=ψF. c and d, The curves were obtained by detecting head turns through thresholded maxima in the total angular velocity of the head (low-pass filtered with a Gaussian with sigma=5 ms to reduce noise). Of the 6697 turns thus detected (from flights of >2 s duration in 4 flies), turns with a total head yaw between 10 and 20 deg to the left were selected. Of the resulting 713 turns, the 100 ms yaw, pitch, and roll surrounding the detection point were averaged; averages for each of the 4 flies separately gave similar results, and movements consistent with this were measured in 13 flies where only head movements were recorded, and 6 flies where only thorax movements were recorded. The mean of each trace was subtracted to cancel offset (such as arbitrary average yaw). Before averaging, ht was obtained for each turn by multiplying the inverse of the t rotation matrix with the h rotation matrix10. Note that ht is not exactly h-t, because ht is defined in a different coordinate system.
Figure 2. a, Full scale corresponds to a probability density 1.2 10-3 deg-1s. b, Full scale corresponds to a probability density 2.4 10-2 deg-1s. For each of the 6697 turns detected (see the legend of Fig. 1), the rise time of the total angular velocity of the head (time between 10% and 90% amplitude) was calculated; the turn was defined as this period, extended by 25% both at onset and end in order to include the early and late phases of both head and thorax turns. All periods thus defined amounted to 37% of the total flight time. The remaining time was designated as 'between turns'.
Rights and permissions
About this article
Cite this article
Schilstra, C., van Hateren, J. Stabilizing gaze in flying blowflies. Nature 395, 654 (1998). https://doi.org/10.1038/27114
Issue Date:
DOI: https://doi.org/10.1038/27114
This article is cited by
-
Two pursuit strategies for a single sensorimotor control task in blowfly
Scientific Reports (2020)
-
Cortical neurons of bats respond best to echoes from nearest targets when listening to natural biosonar multi-echo streams
Scientific Reports (2016)
-
Tracking improves performance of biological collision avoidance models
Biological Cybernetics (2012)
-
Saccadic flight strategy facilitates collision avoidance: closed-loop performance of a cyberfly
Biological Cybernetics (2008)
-
Free flight maneuvers of stalk-eyed flies: do eye-stalks affect aerial turning behavior?
Journal of Comparative Physiology A (2007)
Comments
By submitting a comment you agree to abide by our Terms and Community Guidelines. If you find something abusive or that does not comply with our terms or guidelines please flag it as inappropriate.