1. The anatomy and physiology of a directionally selective motion-detecting (DSMD) neurone in the locust are described. The neurone was descending, with the cell body in the protocerebrum. The axon lay in the dorsolateral quadrant of the nerve cord and has been traced as far as the metathoracic ganglion. It arborized, ipsilateral to the cell body, from the dorsal intermediate tract (DIT) in the suboesophageal and thoracic ganglia.

  2. The neurone was binocular and sensitive to motion in the horizontal plane. It had a preferred direction backwards over the ipsilateral eye and forwards over the contralateral eye. Motion in the opposite direction suppressed the discharge, which had a frequency of 5–20 spikes s−1 at resting membrane potential.

  3. The neurone showed a clear directional response to stimuli with temporal frequencies between 0.7 and 44Hz, with a peak response at 11–22 Hz. It responded with spikes to light ON and light OFF.

  4. The neurone responded directionally to spatial frequencies of 0.28 cycles degree−1 (3.7° stripe period) to above 0.025 cycles degree−1 (40° stripe period). The maximum response was at around 0.035 cycles degree-1 (29° stripe period).

  5. No evidence of adaptation was seen in the responses of the neurone to real or apparent continuous horizontal motion in either the preferred or the null direction.

Neurones in the visual system which are excited by movement in one particular direction (preferred direction) and inhibited by movement in the opposite direction (null direction) have been found in a wide range of invertebrate and vertebrate species. Such directionally selective motion-detecting (DSMD) neurones are thought to underlie an important class of visually guided responses, termed optomotor responses, whereby an image is stabilized in a fixed position on the retina. In insects, a role in optomotor behaviour for descending DSMD neurones (those that have axons descending from the brain) is supported by direct physiological observations in the moth (Rind, 1983a,b), by anatomical and indirect physiological evidence in the fly (Strausfeld and Bassemir, 1985; for a summary of the physiology see Rind, 1983b), and by indirect physiological evidence in the locust (Kien, 1974a,b, 1977).

A mathematical model has been proposed for the types of computational operations which might occur in the central nervous system of a beetle to control optomotor behaviour (Hassenstein and Reichardt, 1956) and has subsequently been found to describe direction-selective motion detection in other insects and in humans (Poggio and Reichardt, 1973; van Santen and Sperling, 1984, 1985). The model involves a correlation between neighbouring input channels. The input signals from two retinal sampling stations are multiplied after the signal from one of the stations has been delayed or low-pass filtered with a characteristic time constant. As a result of these operations, the response of the detector depends not simply on the velocity of the stimulus, but also on the spatial frequency of the stimulus. The response typically shows a maximum for a certain temporal frequency. The determination of this response optimum allows an estimation of the time constant of the low-pass filter. Also, the spatial frequency response of the detector can be used to reveal the spatial organization of the detector inputs. The average response of the detector becomes negative (that is, directional selectivity is reversed and the null direction becomes the preferred) when the spatial wavelength of the stimulus becomes smaller than twice the angular separation of the sampling stations, owing to geometric interference. This reversal has been observed in the response of DSMD neurones in the fly lobula and occurs at the angle separating individual ommatidia, implying that photoreceptors within neighbouring ommatidia form the sampling stations (Eckert, 1973; Buchner, 1984).

The correlation model introduced by Reichardt and coworkers describes the direction-selective process at a phenomenological level, not at the level of synaptic interactions between identified neurones in the visual system (for a review, see Reichardt, 1987). This approach has provided a framework for hunting for specific operations between neurones. In the locust, responses from DSMD neurones have been described at three levels in the visual system: in the second optic neuropile (the medulla, Osorio, 1986), from the stalk of the optic lobe (Kien, 1974b, 1977) and from the circumoesophageal connectives (Kien, 1974b, 1977). Recording from vertically sensitive DSMD neurones in the medulla has shown that the interaction conferring a directional response occurs between adjacent units or channels separated by a single interommatidial angle from one another and depends, at least in part, on an inhibitory interaction between adjacent channels sensitive to luminance change (Osorio, 1986). This supports the proposal of a vetoing operation, proposed by Barlow and Levick (1965) to explain in cellular terms the processes underlying directionally selective motion detection in retinal ganglion cells of the rabbit. In contrast, in the fly eye, by stimulating neighbouring rhabdomeres numbers 1 and 6 within a single ommatidium (and hence two neighbouring cartridges), Riehle and Francheschini (1984) found evidence that the response of the lobula H1 neurone (Hausen, 1976) depends on an excitatory interaction between adjacent channels. The DSMD neurones of the distal locust medulla project to the protocerebrum without arborizing in the lobula. They are localized to the lateral ventral parts of the medulla and are sensitive to vertical movement (Osorio, 1986).

An adaptation in response to continuous movement is shown by all DSMD neurones identified in the fly lobula, and in the cat cortex, and by those DSMD neurones whose responses have been measured indirectly in the human, using psychophysics (Hausen, 1982; Maddess and Laughlin, 1985; Greenlee and Heitger, 1988; Maddess et al. 1988). Maddess and Laughlin (1985) suggest that this adaptation sharpens the response of the neurone to changes in temporal frequency. Adaptation has been shown to shift the time constant of the directionally selective motion-detection process (de Ruyter van Steveninck et al. 1986). The time constant of a motion detector describes the limits of resolvable temporal frequency, and thus any process which has been shown to change the time constant has important functional consequences for the detection of motion (Egelhaaf and Reichardt, 1987; Guo and Reichardt, 1987).

In this paper the morphology of a unique descending DSMD neurone in the locust brain is described. The responses of the neurone to real and apparent movement are used to test the apparent generality of the direction-selective motion-detection process, and also to test the fitness of the Hassenstein and Reichardt correlation model of directionally selective motion detection to describe the direction-selective response of the locust horizontal-motion detectors.

Adult Locusta migratoria were purchased from Animal Magic, Brighton. Experiments were performed at 17–19°C, using two different procedures.

Response characteristics recorded from the axon in the nerve cord

The locust was mounted intact, dorsal side up, and a dorsal midline incision was made along the thorax. The thorax was gently opened, the gut was removed, and a plastic-coated platform was manipulated under the cervical connectives. Recordings were made with glass electrodes, filled with 2 mol 1−1 potassium acetate, which had d.c. resistances in saline of 10–20 MΩ. Recordings were made from the axon of the neurone either intracellularly or just extracellularly, and window circuits were used to discriminate extracellularly recorded action potentials. The locust viewed a television screen (Joyce Electronics, Cambridge) placed either 15 or 30cm from the eye. At 15 cm, the pattern on the screen subtended 90°×60.5° at the eye, and was aligned parallel to the flat sideways-looking ommatidia of one eye, in the equatorial region. The blank screen had a luminance of 160 cd m−2. The pattern viewed during experiments was a sinusoidally modulated luminance profile of maximum luminance 160 cd m−2 and Michelson contrast of 0.5. The profile was produced and controlled by a PDP 11/73 computer and had a frame repetition rate of 200 Hz. The luminance profile (stripes) was aligned vertically, and moved horizontally across the screen at temporal frequencies between 0 and 62 Hz. The spatial frequency of the stripes ranged from 0.025 to 0.8 cycles degree−1. The angular subtense of the pattern could be reduced horizontally or vertically by windowing.

In addition to the Joyce Electronics screen, two arrays of 18 small, green rectangular LEDs, each subtending 5°×8° at the eye, were also used. The LEDs were used to produce apparent movement of a striped pattern. The two arrays were each aligned horizontally over an 8°×90° arc just ventral to the equatorial region of the eye. Within each array, the LEDs were controlled to produce apparent movement of a striped stimulus, each LED within the arc representing a stripe. Six LEDs per array were illuminated simultaneously, each separated by two unilluminated LEDs. Apparent movement consisted of sequential illumination of the six neighbouring LEDs. The sequential illumination of the LEDs was recorded as a stepping trace, each step indicating extinction of six LEDs and simultaneous illumination of their neighbours. The sequential illumination or extinction proceeded from left to right for eight sequences, and then reversed for eight. In all figures, an upward step indicates an anticlockwise apparent movement to the left (backwards over the left eye, forwards over the right) and a downward step indicates a clockwise apparent movement to the right. Apparent movement, such as that produced by LED illumination, has been found to be effective in exciting motion-detecting neurones in the fly (Pick and Buchner, 1979; Riehle and Franceschini, 1984) and, more recently, in the locust (Osorio, 1986; Rind, 1987).

The protocerebral, descending direction-selective motion-detecting (PDDSMD) neurone was identified in these experiments by the position of its axon in the connectives and by its characteristic directionally selective response to apparent horizontal movement over both eyes. In particular, the neurone was identified by its discrete response to each apparent movement in the preferred direction, and its silence when movement was in the null direction. A neurone identified in this way, in a subsequent experiment, was filled via its axon in the cervical connectives and was found to have the same morphology in the brain and a similar axon position in the suboesophageal connectives to those PDDSMD neurones filled and identified after recording from their cell bodies in the brain.

Response characteristics recorded intracellularly from the cell body in the protocerebrum

The locust was dissected and the brain exposed following the method described by Rind (1987). Intracellular recordings were made using glass capillary microelectrodes filled with a saturated solution of hexamminecobaltic chloride. Electrodes had resistances, in saline, of 30–6 0MΩ. All 18 neurones characterized were stained using 15 nA pulses of positive current every second for 1 h. The brains were fixed in 10% formaldehyde buffered to pH 7. Stained neurones were intensified (Bacon and Altman, 1977) and later drawn in whole-mounts of the brain and optic lobe. Some preparations were embedded in Spurr or LEMIX (EMscope) resin. The suboesophageal, prothoracic and mesothoracic ganglia were serially crosssectioned (20 μm sections) to determine in which tract the axon was located, and where the branches projected within each ganglion. In addition, cross-sections, 5 μm thick, were taken of the axon in the suboesophageal connectives. In several preparations, neurones on both sides of the same preparation were characterized and stained.

Apparent movement of a striped pattern was produced by two arrays of 18 LEDs, as described in the previous section. The two arrays were placed and controlled independently of each other. In most experiments, each array was aligned horizontally over an 8°× 90° arc in the equatorial-dorsal region of the eye. The arrays could also be aligned vertically to test sensitivity to vertical apparent motion.

Identification of the neurone: its response to horizontal movement

The protocerebral DSMD neurone was identified in these experiments by the position of either its axon in the cervical connectives or its cell body in the protocerebrum, combined with its characteristic directionally selective response to apparent horizontal movement over both eyes. In particular, the neurone was identified by its discrete response to each apparent movement at a velocity of 90°s− 1 and below in the preferred direction, and by its silence in response to movement in the null direction (Fig. 1). The neurone whose response is shown in Fig. 1 had a cell body in the right protocerebrum. Spikes and EPSPs were recorded in the cell body of the neurone. Depolarizing the neurone using 1–5 nA of current injected at the cell body led to spikes being generated by the EPSPs. In the absence of stimulus movement, the neurones had a resting discharge of 5–20 spikes s−1. Apparent movement in the preferred direction (Fig. 1 backwards over the ipsilateral, forwards over the contralateral eye) increased this resting discharge, whereas movement in the null direction (Fig. 1 forwards over the ipsilateral, backwards over the contralateral eye) suppressed the discharge. The excitatory input in the preferred direction and the suppression in the null direction were received from both eyes (Fig. 1C,D). At 17–19°C this response had a latency of 65–80 ms from the apparent movement step to the first upward inflection in the response at the neurone. No IPSPs were observed in response to movement in the null direction over either eye, even when the neurone was depolarized by injected positive current to accentuate IPSPs. The response of the neurone in the preferred direction did not wane over the course of the experiment. Horizontal directions of movement produced much stronger responses than those produced by vertical movements.

Fig. 1.

Response to movement of a visual stimulus, recorded intracellularly from the cell body of the PDDSMD neurone in the right protocerebrum. The response of the ipsilateral cell (after Rind, 1987) was monitored indirectly via an extracellular recording from the axon of the DSMD neurone in the cervical connectives (arrow in traces A,B,D). The non-directional LGMD1 is not excited by stimuli which excite PDDSMD. (A) Resting discharge in the absence of apparent movement; (B-D) responses of the neurone to apparent horizontal movement at 180°s−1 to the right (downward stepping on bottom trace) and the left (upward stepping on bottom trace) (B) over both eyes, (C) over the right eye and (D) over the left eye.

Fig. 1.

Response to movement of a visual stimulus, recorded intracellularly from the cell body of the PDDSMD neurone in the right protocerebrum. The response of the ipsilateral cell (after Rind, 1987) was monitored indirectly via an extracellular recording from the axon of the DSMD neurone in the cervical connectives (arrow in traces A,B,D). The non-directional LGMD1 is not excited by stimuli which excite PDDSMD. (A) Resting discharge in the absence of apparent movement; (B-D) responses of the neurone to apparent horizontal movement at 180°s−1 to the right (downward stepping on bottom trace) and the left (upward stepping on bottom trace) (B) over both eyes, (C) over the right eye and (D) over the left eye.

On 17 separate occasions, neurones with the response characteristics described above were stained intracellularly, and the following morphological characteristics were consistently revealed. In one preparation (Fig. 2) the neurones on both sides of the same preparation were filled. It was concluded from the combined morphological and physiological evidence that there was only one such neurone on each side of the animal.

Fig. 2.

Morphology of the left and right PDDSMD neurones in the brain. (A) Camera lucida drawings from whole-mounts of two neurones stained in the same preparation. The brain is viewed from behind. (B) Cross-section through the circumoesophageal connective showing the location of the stained axon (arrowhead) of the right directionally selective motion-detecting neurone drawn in A.

Fig. 2.

Morphology of the left and right PDDSMD neurones in the brain. (A) Camera lucida drawings from whole-mounts of two neurones stained in the same preparation. The brain is viewed from behind. (B) Cross-section through the circumoesophageal connective showing the location of the stained axon (arrowhead) of the right directionally selective motion-detecting neurone drawn in A.

Identification of the neurone: its morphology

The 25–30 μm diameter cell body of the protocerebral, descending direction-selective motion-detecting neurone (PDDSMD neurone) was located in a cortex of cell bodies on the posterior slope of the protocerebrum. The cell body lay within the V formed by the tracheae entering the posterior slope at the protocerebrum, 50 gm ventral and 30 μm more medial to the conspicuous DCMD (O’Shea et al. 1974) cell body (Figs 2, 3). A 5–10 μm diameter neurite emerged from the cell body and projected ventrally for 90 μm, before giving off a series of five branches which projected medially for 100gm. No branches crossed the midline of the brain. The first large (>7 μm) process projected laterally for 200 μm towards the ipsilateral optic lobe. A second, and sometimes a third, large process emerged from the neurite 20–50 μm after the first, and projected both laterally towards the ipsilateral optic lobe and anterioventrally towards the deuterocerebrum. These anterior processes did not enter the antennal lobe of the deuterocerebrum. The main process of the neurone narrowed slightly after the emergence of the second (or third, if present) large branch, and bent 25° towards the midline of the brain before producing 5–7, 1 gm diameter processes in the tritocerebrum, which projected 20–30 μm from the axon of the neurone. These processes were thicker than the processes of the neurone in the protocerebrum. The 10–15 μm diameter axon of the neurone then projected to a position in the dorsolateral quadrant of the ipsilateral connective (Fig. 2B). In Fig. 2, the left and right neurones have been filled in the same preparation. The inset shows a 5 μm diameter section through the right connective, at the level indicated by the arrow. The stained profile of the neurone is indicated by an arrowhead. The dorsally located DCMD axon is the largest profile in the nerve cord. The PDDSMD neurone has been filled from the cervical connectives as far as its projections into the mesothoracic ganglion (Fig. 3). Its branches in the suboesophageal, pro- and mesothoracic ganglia were all restricted to the dorsal, ipsilateral half of the ganglion (Fig. 3). The axon projected through the suboesophageal, prothoracic and mesothoracic ganglia in the lateral, dorsal quadrant of the dorsal intermediate tract (DIT) (Tyrer and Gregory, 1982). The mediolateral extent of the arborizations in the the ipsilateral connective (Fig. 2B). In Fig. 2, the left and right neurones have been filled in the same preparation. The inset shows a 5 pm diameter section through the right connective, at the level indicated by the arrow. The stained profile of the neurone is indicated by an arrowhead. The dorsally located DCMD axon is the largest profile in the nerve cord. The PDDSMD neurone has been filled from the cervical connectives as far as its projections into the mesothoracic ganglion (Fig. 3). Its branches in the suboesophageal, pro- and mesothoracic ganglia were all restricted to the dorsal, ipsilateral half of the ganglion (Fig. 3). The axon projected through the suboesophageal, prothoracic and mesothoracic ganglia in the lateral, dorsal quadrant of the dorsal intermediate tract (DIT) (Tyrer and Gregory, 1982). The mediolateral extent of the arborizations in the suboesophageal, pro- and mesothoracic ganglia exceeded the dorsoventral extent by a ratio of 3:1.

Fig. 3.

Morphology of the left PDDSMD neurone in (A) the brain, and (B) the suboesophageal, (C) the prothoracic and (D) the mesothoracic ganglia. Camera lucida drawings from whole-mounts of brains viewed either (A) from behind or (B–D) dorsally. Scale bars, 100 μm (A), 200 μm (B–D).

Fig. 3.

Morphology of the left PDDSMD neurone in (A) the brain, and (B) the suboesophageal, (C) the prothoracic and (D) the mesothoracic ganglia. Camera lucida drawings from whole-mounts of brains viewed either (A) from behind or (B–D) dorsally. Scale bars, 100 μm (A), 200 μm (B–D).

Temporal frequency response

To quantify the response of the PDDSMD neurone to real movements, spikes were recorded from its axon in the cervical connectives in locusts with intact heads. A pattern subtending 90°×60.5° at the eye could be made to move at various velocities across a television screen, viewed by one of the locust’s eyes. First, the optimum spatial frequency was selected for each neurone. Next, the responses (spikes s−1) to stripe movements at a series of seven different temporal frequencies were recorded (Fig. 4). Each temporal frequency was presented 10 times, in random order (that is N=10). Temporal frequencies in the range 0.71–5.5 Hz were presented for 2.5 s and the mean spike rate sampled at 10-ms intervals over 1.75 s. Temporal frequencies between 7.8 and 62.4Hz were presented for Is and the mean spike rate sampled at 10-ms intervals over a 0.5-s period. The mean responses of the neurone during the sampling period (and standard error of the mean) were plotted against temporal frequency (Fig. 4A,B). The response to a temporal frequency of 5.5 Hz measured throughout the presentation time of the stimulus is shown for motion in the preferred (Fig. 4Ci) or null directions (Fig. 4Cii) or in the absence of movement (Fig. 4Ciii). There was no modulation of the response to movement in either the preferred or null directions at any of the temporal frequencies tested. The peak in the response at temporal frequencies of 11–22 Hz and the lack of a directional response above 44 Hz were consistent in seven preparations. The response was consistent between the two neurones on either side of an animal. A clear directional response was given at the lowest temporal frequencies tested (0.7Hz).

Fig. 4.

Temporal frequency response of the PDDSMD neurone. The mean spike rate is plotted as a function of temporal frequency. Stimuli of optimal spatial frequency for this neurone (period of 29°) were used. Each temporal frequency was presented 10 times in random order. The stimuli were presented as two series of seven temporal frequencies. (A) A low temporal frequency series in which each frequency was presented for 2.5 s and spike rate sampled over a 1.75-s period. Standard errors are also shown. (B) A high temporal frequency series in which each frequency was presented for Is and spike rate sampled for 0.5s. Even at the lowest temporal frequencies the locust viewed an entire luminance profile cycle. As shown in C for a temporal frequency of 5.5 Hz, the response of the neurone is unmodulated. Each bin is 10 ms in width and represents the total number of spikes s−1 for the 10 presentations of each temporal frequency. Recordings were made from the axon of the neurone in the cervical connectives.

Fig. 4.

Temporal frequency response of the PDDSMD neurone. The mean spike rate is plotted as a function of temporal frequency. Stimuli of optimal spatial frequency for this neurone (period of 29°) were used. Each temporal frequency was presented 10 times in random order. The stimuli were presented as two series of seven temporal frequencies. (A) A low temporal frequency series in which each frequency was presented for 2.5 s and spike rate sampled over a 1.75-s period. Standard errors are also shown. (B) A high temporal frequency series in which each frequency was presented for Is and spike rate sampled for 0.5s. Even at the lowest temporal frequencies the locust viewed an entire luminance profile cycle. As shown in C for a temporal frequency of 5.5 Hz, the response of the neurone is unmodulated. Each bin is 10 ms in width and represents the total number of spikes s−1 for the 10 presentations of each temporal frequency. Recordings were made from the axon of the neurone in the cervical connectives.

Spatial frequency response

The responses to a series of different spatial frequencies of pattern (0.025–0.8 cycles degree−1) were recorded from the PDSMD axon in the cervical connectives (Figs 5, 6). The selected velocity of pattern movement was 15.6Hz, at which the neurone gave an optimum response in the preferred direction. Patterns were moved in either the preferred or the null direction for 0.5 s. Each separate exposure to one spatial frequency was given 10 times in random order. The subtense of the pattern on the screen could be varied, as could the position of the screen relative to the locust’s eye. The screen was either 15 cm (Fig. 5) or 30cm (Fig. 6) from the eye. Thus, the range of spatial frequencies was 0.025–0.4 cycles degree−1 (Fig. 5) or 0.05–0.8 cycles degree−1 (Fig. 6). The response of the neurone to the largest pattern at 15 cm from the eye is given in Fig. 5A. The response in the preferred direction increased as the spatial frequency of the pattern decreased, with a maximum at spatial frequencies of 0.035 cycles degree−1 (i.e. 29° stripe period). There was no clear directional response to spatial frequencies of 0.4 cycles degree−1 (i.e. 2.5° stripe period) and above. Reducing the amount of pattern shown (Fig. 5B,C) lowered the response of the neurone to pattern movement. It also revealed an apparent crossing over of responses to movement in the preferred and null directions at spatial frequencies around 0.28 cycles degree−1 (i.e. 3.7° stripe period). This apparent crossing over was investigated further by increasing the distance from the eye to the screen (from 15 to 30 cm) and, hence, changing the range of spatial frequencies (Fig. 6). When this was done, the spatial frequency at which there was no clear directional response was 0.4 cycles degree−1 (i.e. 2.5° stripe period). At 0.56 cycles degree−1 (1.9° stripe period) the response to motion in the preferred direction and null direction also appeared to reverse.

Fig. 5.

Spatial frequency response of the PDDSMD neurone. Mean spike rate is plotted as a function of spatial frequency (cycles degree−1). The pattern subtense over one eye was (A) 90°×60.5° (B) 90°×8° or (C) 45°×8°. The pattern had a temporal frequency of 15.6Hz (the optimum for this neurone; see Fig. 6 from the same neurone). Each point is the mean response to 10 presentations of each spatial frequency for 0.5 s and given in a random order. Standard errors (which were small) are also shown.

Fig. 5.

Spatial frequency response of the PDDSMD neurone. Mean spike rate is plotted as a function of spatial frequency (cycles degree−1). The pattern subtense over one eye was (A) 90°×60.5° (B) 90°×8° or (C) 45°×8°. The pattern had a temporal frequency of 15.6Hz (the optimum for this neurone; see Fig. 6 from the same neurone). Each point is the mean response to 10 presentations of each spatial frequency for 0.5 s and given in a random order. Standard errors (which were small) are also shown.

Fig. 6.

Spatial frequency response of the PDDSMD neurone. Mean spike rate is plotted against spatial frequency (cycles degree−1). The pattern subtended was 53°8′×32°42′ over one eye and had a temporal frequency of 15.6Hz. See text for further details of stimulus presentation. Standard errors are shown.

Fig. 6.

Spatial frequency response of the PDDSMD neurone. Mean spike rate is plotted against spatial frequency (cycles degree−1). The pattern subtended was 53°8′×32°42′ over one eye and had a temporal frequency of 15.6Hz. See text for further details of stimulus presentation. Standard errors are shown.

Adaptation of the response

Responses recorded intracellularly from the cell body in the brain, or from the axon in the neck in a minimally dissected locust, showed no sign of adaptation to maintained moving stimuli. Fig. 7A,B shows the response of PDDSMD in a minimally dissected locust to real movement at speeds of 450° s−1 (16 Hz) using a sinusoidally modulated luminance profile (stripe period) with a spatial period of 29° (an optimum period for this neurone) viewed on a Joyce screen. The pattern was a bright (160 cd m−2) high-contrast (0.5) pattern subtending 55°×40° over one eye. First, the optimal spatial and temporal frequency of the pattern were determined for each neurone. Then followed three possible 20-s blocks of stimulus. Each of the three possible blocks was repeated 10 times in random sequence. The initial 5 s of the 20-s block consisted of one of three adapting stimuli: (1) the optimal stimulus in the preferred direction; (2) the optimal stimulus moving in the null direction; (3) a blank screen. Each of these 5-s adapting exposures was immediately followed by a 5-s test exposure to the optimal stimulus moving in the preferred direction, then by a 10-s recovery period before the next adapting stimulus was delivered. The response was plotted during each 0.5 s of the 5-s test exposure, and a comparison was made between the responses following the three possible adapting stimuli. The first data point in each graph was taken from the last 0.5 s of the adapting stimulus. There was neither a clear trend in firing rate during the course of the test stimulus, nor any consistent difference in the response to the test stimulus following the adapting stimuli. During these experiments, a clear ON response was produced by changing from a blank screen to either the null or the excitatory stimulus. This response accounts for the high response to the first 0.5 s of the test stimulus following the blank-screen adapting stimulus. Fig. 8 shows the intracellularly recorded response to 18 s of continuous apparent movement, at 90°s−1, of a pattern subtending 8°×90° over each eye.

Fig. 7.

(A,B) Lack of adaptation in the response of PDDSMD. Spike rate is plotted against time during 5 s of optimal stimulus in the preferred direction, which followed 5 s of optimal stimulus in the preferred direction, or in the null direction or a blank screen. The first data point was from the last 0.5 s of the adapting stimulus. A 10-s rest occurred between each trial. (A) Recordings were made from the axon of the PDDSMD neurone in the left connective. Figs 4 and 5 show temporal and spatial frequency plots using data from the same neurone. (B) Recordings were made from the axon of the PDDSMD neurone in the right connective of the preparation. Standard errors are shown.

Fig. 7.

(A,B) Lack of adaptation in the response of PDDSMD. Spike rate is plotted against time during 5 s of optimal stimulus in the preferred direction, which followed 5 s of optimal stimulus in the preferred direction, or in the null direction or a blank screen. The first data point was from the last 0.5 s of the adapting stimulus. A 10-s rest occurred between each trial. (A) Recordings were made from the axon of the PDDSMD neurone in the left connective. Figs 4 and 5 show temporal and spatial frequency plots using data from the same neurone. (B) Recordings were made from the axon of the PDDSMD neurone in the right connective of the preparation. Standard errors are shown.

Fig. 8.

Lack of adaptation of the intracellularly recorded response to maintained apparent movement, first in the preferred and then in the null direction. Records A-C are part of a continuous recording; 12 s separate the end of A from the start of B, and a further 12 s separates B from C. The resting discharge was recorded in the absence of stimulus movement. The recording is from the cell body of the left DSMD neurone. Upward stepping on the lower trace represents apparent movement at 90°s−1, to the left, over both eyes. These records are from the same neurone as those in Fig. 4.

Fig. 8.

Lack of adaptation of the intracellularly recorded response to maintained apparent movement, first in the preferred and then in the null direction. Records A-C are part of a continuous recording; 12 s separate the end of A from the start of B, and a further 12 s separates B from C. The resting discharge was recorded in the absence of stimulus movement. The recording is from the cell body of the left DSMD neurone. Upward stepping on the lower trace represents apparent movement at 90°s−1, to the left, over both eyes. These records are from the same neurone as those in Fig. 4.

The interpretation of the intracellularly recorded response to apparent movement is complex. In the preferred direction, PDDSMD is responding to each apparent motion step rather than to continuous motion, whereas in the null direction the resting discharge is continuously suppressed. Clearly the response in the null direction does not adapt.

The protocerebral, descending DSMD (PDDSMD) neurone has not been identified previously. The cell body is a member of the protocerebral posterior pars intercerebralis group of Williams (1975), although it has not been described individually. Kien (1974b, 1977) recorded extracellularly from DSMD neurones (termed Bl and B2) with axons in the circumoesophageal connectives. Like PDDSMD, these neurones gave a directional response to horizontal movement, with the left neurone having a preferred direction backwards over the ipsilateral eye and forwards over the contralateral one, and the right neurone having the opposite preferred directions. Kien concluded that neurones with response properties matching those of the descending DSMD neurones Bl and B2 produced the optomotor torque response monitored from neck muscles (50 and 51) and from the motoneurones innervating them. Like Bl, B2 and PDDSMD, the fast and slow motoneurones innervating the neck muscles respond to velocities of movement of 475°s−1 and above. In Kien’s study this represents a temporal frequency of 25 Hz, because a 19° stripe period was used. The optimum temporal frequency cannot be estimated from the data of Kien, because the response did not peak and was still increasing at temporal frequencies of 25Hz. Neurones like Bl, B2 and PDDSMD could contribute to the optomotor response in the horizontal (yaw) plane.

Other motion-sensitive visual interneurones in the locust include the ‘deviation detectors’. Identified deviation-sensitive neurones DNM, DNI, DNC, TCG and Pl(2)5 are excited by specific combinations of sensory modalities, all signalling deviation from a straight flight path (Simmons, 1980; Mohl and Bacon, 1983; Reichert et al. 1985; Hensler, 1988). DNI, DNC and Pl(2)5 are all protocerebral, descending neurones and, like PDDSMD, they are direction-selective and are sensitive to horizontal movements over the compound eyes (Reichert et al. 1985 ; Hensler, 1988). The axons of these descending neurones project in distinct tracts in the thoracic nervous system. PDDSMD and TCG have axons in the DIT tract (Bacon and Tyrer, 1978), whereas DNI, DNM, DNC and Pl(2)5 project in a separate tract – MDT (Gris and Rowell, 1986; Hensler, 1988).

Such descending neurones influence motoneurones in two ways. In the moth, for example, descending DSMD neurones contribute directly to optomotor responses by monosynaptic excitation of motoneurones to specific flight muscles (Rind, 1983b). However, multimodal deviation-sensitive neurones in the locust, such as DNM, DNI and DNC, make both direct connections with specific motoneurones to flight muscles capable of producing a turn during flight (Simmons, 1980) and indirect connections via premotor interneurones. It has been suggested that the indirect connections act in flight, during each wing beat cycle, to alter activity in flight motoneurones capable of correcting a perceived deviation from a straight flight path (Reichert et al. 1985).

Spatial frequency response

In the present investigations, the DSMD neurone gave a directional response to spatial frequencies down to periods of between 2.5° and 3.7°, values equal to twice the interommatidial angle in this region (Horridge, 1978). Light-adapted locust photoreceptors in the equatorial region of the eye have acceptance angles of 1.4° (width of the sensitivity function at half of maximum; Wilson, 1975). This suggests that the direction-selective motion-detection process can occur between sampling bases beneath adjacent ommatidia. The flat screen upon which the stimuli were presented will introduce a distortion factor. From an angle of 45°, a 3° stripe will have an apparent width of 2°. Increasing the distance of the screen from the locust (Fig. 6) or decreasing the extent of the pattern (Fig. 5B,C) minimizes this distortion but does not change the observed limit of spatial resolution. Furthermore, the experimental results suggest a crossover in the directional response (the preferred direction becomes the null direction) to movements of patterns with a period of around 3.7°. This was particularly clear when the extent of the pattern was reduced, lessening the pattern distortion both because of gradual changes in horizontal alignment of rows of ommatidia over the eye and because of a decrease in the apparent spatial frequency of a pattern with increasing viewing angle of the stimulus from the eye (Figs 5B,C, 6). The crossover in response was not statistically significant but was observed consistently. Further experiments are necessary to confirm the crossover statistically.

Temporal frequency response, time constants and adaptation

The PDDSMD neurone shows a response maximum which depends on the temporal frequency of the stimulus (Fig. 4A,B). This indicates that it receives input from motion detectors of the correlation-type and not from detectors using a different algorithm, leading to the functional characterization of the directionally selective motion detectors in the locust visual system as correlation motion detectors. Furthermore, the unmodulated response (Enroth-Cugell and Robson, 1966) of PDDSMD suggests that a non-linear process, such as the multiplicative stage of the Reichardt correlation motion detector, underlies the frequency response of PDDSMD.

The optimum temporal frequency, for a pattern repeat of 29°, was 15.6Hz. The neurone responded directionally to temporal frequencies ranging from around 31 Hz down to 0.75 Hz. In flies, the optimum temporal frequencies have been estimated for DSMD neurones (H1 and HSE) in the lobula, for the optomotor response and also for the behavioural landing response. The responses of lobula neurones H1 and HSE and the optomotor response show temporal frequency optima in the range 1–10 Hz (Gotz, 1964; Eckert, 1973, 1980; Mastebroek et al. 1980; Hausen, 1982; Buchner, 1984; Maddess and Laughlin, 1985). Maddess and Laughlin (1985), Egelhaaf and Reichardt (1987) and Guo and Reichardt (1987) suggest that this range is an underestimate of the optima, particularly because a time-averaged response was taken so that any process of adaptation would be ignored. The unadapted temporal frequency optimum, when measured, is at 8–10 Hz (H1, Maddess and Laughlin, 1985; HSN HSE, HSS, Hausen, 1982). This value is close to the temporal frequency optimum of the landing response measured as the shortest latency to landing (Borst and Bahde, 1986). By comparison, the optimum of the locust PDDSMD neurone response occurs at higher temporal frequencies, suggesting that the filter time constant of the directionally selective motion-detecting process is shorter, in the locust, than that measured for either the landing response or for direction-selective motion detection in the fly. Recently, the time constant of the direction-selective motiondetecting process has been measured both behaviourally and from the response of lobula DSMD neurones without using the temporal frequency optimum. The time constants of the processes underlying the behavioural response and of the response of the lobula neurones were found to be similar, and were less than 20 ms (Egelhaaf and Reichardt, 1987; Guo and Reichardt, 1987). The time constant of the landing response is 10–30 ms (Borst and Bahde, 1986). Egelhaaf and Reichardt (1987) state that temporal frequency optima can only be used as a rough guide to the time constant of the directionally selective motion-detecting process as they are typically broad about their peak. However, they give a range of observed temporal frequency optima between 9 and 90 ms. The time constant of the directionally selective motion-detecting process in the locust, estimated from the optimal temporal frequency response of the PDDSMD neurone following Egelhaaf and Reichardt (1987), would be shorter than these estimates (7.5 ms). This may be an underestimate of the filter time constant, as the contrast modulation of the stimulus (0.5) may have introduced a saturating non-linearity into the process. In the fly, introducing such a non-linearity leads to a lower estimate of the filter time constant, but only when it is combined with velocities of motion above about 10° s−1 (15 Hz Fig. 3; de Ruyter van Steveninck et al. 1986). In summary, this means that the locust PDDSMD neurone is tuned to higher frequencies than are fly lobula plate cells, a finding which is particularly noteworthy given the comparatively slow impulse response of the locust retinula cells (Howard et al. 1984).

One of the most striking features of the response of the PDDSMD neurone is its lack of adaptation to continuous motion, at high photometric contrasts, over a wide area of the visual field and at high temporal frequencies. These conditions would be expected to accentuate any adaptation. This lack of adaptation may be partly responsible for the short time constant, estimated for the DSMD process from the temporal frequency optimum, because the optimum frequency of the temporal frequency response reduces during adaptation (H1, Maddess and Laughlin, 1985; HSN HSS and HSE, Hausen, 1982). However, the temporal resolution of H1, measured by its ability to follow sudden movements (velocity contrasts), has been found to increase with adaptation (Maddess and Laughlin, 1985; de Ruyter van Steveninck et al. 1986). Adaptation is a general feature of the response of the DSMD neurone in the fly eye. It has been suggested that the adaptation process, which may be governed by temporal frequency, increases the neurone’s ability to signal changes in temporal frequency (Maddess and Laughlin, 1985). Adaptation to a continuous stimulus in the fly lobula motion detector, H1, also shifts the time constant (Maddess and Laughlin, 1985; de Ruyter van Steveninck et al. 1986), as does temporal modulation of luminance without motion (Borst and Egelhaaf, 1987). In the locust PDDSMD neurone, which was found not to adapt, it is not clear how the time constant could be shifted. This raises the question of the nature of a direction-selective motion-detecting process which does not adapt and which cannot, therefore, use a changed time constant to signal contrasts in the velocity of motion. The lack of adaptation shown by the locust descending DSMD neurone may be correlated with the lower overall spike rates (50–60 spikes s−1 optimal response to movement in the preferred direction, cf. 300spikess−1 in H1: Maddess and Laughlin, 1985). In other words, in the experiments reported here the locust PDDSMD may not have been sufficiently excited to cause any adaptation. This explanation seems unlikely, as the size of the screen used to stimulate the DSMD neurone with real movement was much larger than that used by Maddess and Laughlin (1985) to demonstrate adaptation in H1 (90°x60° compared with 58° diameter). In addition, fly lobula plate neurones (HSN, HSE, HSS) also receive binocular inputs and show adaptation when stimulated using a single monocular 36.9° diameter pattern (Hausen, 1982).

I am grateful to Andrew Derrington in whose laboratory most of the experiments were conducted and the data processed. Tracey Pickering helped to perform the experiments. Peter Simmons read the manuscript critically and gave help and support during all facets of the work. I was supported by an SERC Advanced Research Fellowship and an SERC Project Grant.

Bacon
,
J. P.
and
Altman
,
J. S.
(
1977
).
A silver intensification method for cobalt-filled neurones in wholemount preparations
.
Cell and Tissue Res
.
178
,
199
219
.
Bacon
,
J.
and
Tyrer
,
M.
(
1978
).
The tritocerebral commissure giant (TCG): A bimodal interneurone in the locust, Schistocerca gregaria
.
J. comp. Physiol
.
126
,
317
325
.
Barlow
,
H. B.
and
Levick
,
W. R.
(
1965
).
The mechanism of directionally selective units in rabbit’s retina
.
J. Physiol., Lond
.
178
,
477
504
.
Borst
,
A.
and
Bahde
,
S.
(
1986
).
What kind of movement detector is triggering the landing response of the housefly?
Biol. Cybernetics
55
,
59
69
.
Borst
,
A.
and
Egelhaaf
,
M.
(
1987
).
Temporal modulation of luminance adapts time constant of fly movement detectors
.
Biol. Cybernetics
55
,
59
215
.
Buchner
,
E.
(
1984
).
Behavioural analysis of spatial vision of insects
.
In Photoreception and Vision in Invertebrates
(ed.
M. A.
Ali
), pp.
561
622
.
New York, London
:
Plenum Press
.
De Ruyter Van Steveninck
,
R. R.
,
Zaagman
,
W. H.
and
Mastebroek
,
H. A. K.
(
1986
).
Adaptation of transient responses of a movement-sensitive neuron in the visual system of the blowfly Calliphora erythrocephala
.
Biol. Cybernetics
54
,
223
236
.
Eckert
,
H.
(
1973
).
Optomotorische Untersuchungen am visuellen System der Stubenfliege Musca domestica L
.
Kybemetik
14
,
1
23
.
Eckert
,
H.
(
1980
).
Orientation sensitivity of the visual movement detection system activating the landing response of the blowflies, Calliphora and Phoenicia: a behavioural investigation
.
Biol. Cybernetics
37
,
235
247
.
Egelhaaf
,
M.
and
Reichardt
,
W.
(
1987
).
Dynamic response properties of movement detectors: theoretical analysis and electrophysiological investigation in the visual system of the fly
.
Biol. Cybernetics
56
,
69
87
.
Enroth-Cugell
,
C.
and
Robson
,
J. G.
(
1966
).
The contrast sensitivity of retinal ganglion cells of the cat
.
J. Physiol., Lond
.
187
,
517
552
.
Gotz
,
K. G.
(
1964
).
Optomotorische Untersuchung des visuellen Systems einiger Augenmutanten der Fruchtfliege Drosophilia
.
Kybemetik
2
,
77
92
.
Greenlee
,
M. W.
and
Heitger
,
F.
(
1988
).
The functional role of contrast adaptation
.
Vision Res
.
28
,
791
797
.
Gris
,
C.
and
Rowell
,
C. H. F.
(
1986
).
Three descending interneurons reporting deviation from course in the locust
.
J. comp. Physiol
.
158
,
765
774
.
Guo
,
A.
and
Reichardt
,
W.
(
1987
).
An estimation of the time constant of movement detectors
.
Naturwissenschaften
74s
,
91
.
Hassenstein
,
B.
and
Reichardt
,
W.
(
1956
).
System theoretische Analyse der Zeit - Reihenfolgen - und Vorzeichenauswertung bei der Bewegungsperzeption des Riisselkhfers Chlorophanus
.
Z. Naturforsch
.
116
,
513
524
.
Hausen
,
K.
(
1976
).
Functional characterisation and anatomical identification of motion sensitive neurons in the lobula plate of the blowfly Calliphora erythrocephala
.
Z. Naturforsch
.
31c
,
629
633
.
Hausen
,
K.
(
1982
).
Motion sensitive interneurons in the optomotor system of the fly. II. The horizontal cells: receptive field organisation and response characteristics
.
Biol. Cybernetics
45
,
143
156
.
Hensler
,
K.
(
1988
).
The pars intercerebralis neurone PI(2)5 of locusts: convergent processing of inputs reporting head movements and deviations from straight flight
.
J. exp. Biol
.
140
,
511
533
.
Horridge
,
G. A.
(
1978
).
The separation of visual axes in apposition compound eyes
.
Phil. Trans. R. Soc. Ser. B
285
,
1
59
.
Howard
,
J.
,
Dubs
,
A.
and
Payne
,
R.
(
1984
).
The dynamics of phototransduction in insects. A comparative study
.
J. comp. Physiol
.
154
,
707
718
.
Kien
,
J.
(
1974a
).
Sensory integration in the locust optomotor system. I. Behavioural analysis
.
Vision Res
.
14
,
1245
1254
.
Kien
,
J.
(
1974b
).
Sensory integration in the locust optomotor system, n. Direction selective neurons in the circumoesophageal connectives and the optic lobe
.
Vision Res
.
14
,
1255
1268
.
Kien
,
J.
(
1977
).
Comparison of sensory input with motor output in the locust optomotor system
.
J. comp. Physiol
.
113
,
161
179
.
Maddess
,
T.
and
Laughlin
,
S. B.
(
1985
).
Adaptation of the motion-sensitive neuron Hl is generated locally and governed by contrast frequency
.
Proc. R. Soc. B
225
,
251
275
.
Maddess
,
T.
,
Mccourt
,
M. E.
,
Blakeslee
,
B.
and
Cunningham
,
R. B.
(
1988
).
Factors governing the adaptation of cells in area-17 of the visual cortex
.
Biol. Cybernetics
59
,
229
236
.
Mastebroek
,
H. A. K.
,
Zaagman
,
W. H. Z.
and
Lenting
,
B. P. M.
(
1980
).
Movement detection performance of a wide-field element in the visual system of the blowfly
.
Vision Res
.
20
,
467
474
.
Mohl
,
B.
and
Bacon
,
J.
(
1983
).
The tritocerebral commissure giant (TCG) wind-sensitive interneurone in the locust, n. Directional sensitivity and role in flight stabilisation
.
J. comp. Physiol
.
150
,
453
464
.
O’shea
,
M.
,
Rowell
,
C. H. F.
and
Williams
,
J. L. D.
(
1974
).
The anatomy of a locust visual interneurone: the descending contralateral movement detector
.
J. exp. Biol
.
60
,
1
12
.
Osorio
,
D.
(
1986
).
Directionally selective cells in the locust medulla
.
J. comp. Physiol
.
159
,
841
847
.
Pick
,
B.
and
Buchner
,
E.
(
1979
).
Visual movement detection under light- and dark-adaptation in the fly Musca domestica
.
J. comp. Physiol
.
134
,
45
54
.
Poggio
,
T.
and
Reichardt
,
W.
(
1973
).
Considerations on models of movement detection
.
Kybemetik
13
,
223
227
.
Reichardt
,
W.
(
1987
).
Evaluation of optical motion information by movement detectors
.
J. comp. Physiol
.
161
,
533
547
.
Reichert
,
H.
,
Rowell
,
C. H. F.
and
Gris
,
C.
(
1985
).
Course correction circuitry translates feature detection into behavioural action in locusts
.
Nature, Lond
.
315
,
142
144
.
Riehle
,
A.
and
Franceschini
,
N.
(
1984
).
Motion detection in flies: Parametric control over on-off pathways
.
Expl Brain Res
.
54
,
390
394
.
Rind
,
F. C.
(
1983a
).
A directionally sensitive motion detecting neurone in the brain of a moth
.
J. exp. Biol
.
102
,
253
271
.
Rind
,
F. C.
(
1983b
).
The role of an identified brain neurone in mediating optomotor movements in a moth
.
J. exp. Biol
.
102
,
273
284
.
Rind
,
F. C.
(
1987
).
Non-directional, movement sensitive neurones of the locust optic lobe
.
J. comp. Physiol
.
161
,
477
494
.
Simmons
,
P. J.
(
1980
).
A locust wind and ocellar brain neurone
.
J. exp. Biol
.
86
,
87
98
.
Strausfeld
,
N. J.
and
Bassemir
,
U. K.
(
1985
).
The organisation of giant horizontal-motionsensitive neurons and their synaptic relationships in the lateral deutocerebrum of Calliphora erythrocephala and Musca domestica
.
Cell Tissue Res
.
242
,
531
550
.
Tyrer
,
N. M.
and
Gregory
,
G. E.
(
1982
).
A guide to the neuroanatomy of locust suboesophageal and thoracic ganglia
.
Phil. Trans. R. Soc. Ser. B
297
,
91
123
.
Van Santen
,
J. P. H.
and
Sperling
,
G.
(
1984
).
A temporal introduction to the central complex in the locust, Schistocerca gregaria (Orthoptera)
.
J. Zool. Lond
.
176
,
67
86
.
Van Santen
,
J. P. H.
and
Sperling
,
G.
(
1985
).
Elaborated Reichardt detectors
.
J. opt. Soc. Am
.
2
,
300
320
.
Williams
,
J. L. D.
(
1975
).
Anatomical studies of the insect central nervous system: a ground plan of the midbrain and an introduction to the central complex in the locust, Schistocerca gregaria (Orthoptera)
.
J. Zool., Lond
.
176
,
67
86
.
Wilson
,
M.
(
1975
).
Angular sensitivity of dark and light adapted locust retinula cells
.
J. comp. Physiol
.
97
,
323
328
.