The Gregarious Behaviour of Marching Locusta Migratoria Migratorioides (R. & F.) Hoppers

Locusta migratoria migratorioides (R. & F.) is a highly polymorphic species, the differences in anatomy and colour being so great that the extremes were at one time classed as separate species. Variations in behaviour are also striking and for the nymphs or hoppers seem to be related to the density of the population, although this may not be the only factor operating. There are practically no published observations on the behaviour of L. migratoria migratorioides in the field. Information on other locusts has been freely drawn upon, although the behaviour of the various species is probably not identical.

The evidence at present available suggests that, when populations are thin, the hoppers live in relative isolation and spend most of the day basking on the ground or resting in the plants. When hopper populations are dense, the individuals live in groups or bands and carry out a characteristic daytime wandering called marching. Locusts tend to live in relatively uncultivated country, so that their ability to emigrate from these areas into others of intensive cultivation greatly adds to the economic importance of their capacity for very rapid population increases. Although they cannot move as rapidly or as far as adults, yet hopper bands (which may cover many acres) can severely damage crops when they march through cultivated fields. In addition to its economic importance marching behaviour is, like all mass movements of animals, of considerable biological interest.

The individual hoppers of a marching band tend to move in the same general direction, progressing by periods of steady walking, interspersed with long low hops. It appears that in other species of locusts, the hoppers from such bands are far less active when alone (Fraenkel, 1929; Kennedy, 1939, on Schistocerca gregaria (Forsk.)). Clark (1949) gave a good example of the increased marching of large bands. On one particular day a large and a small band of Chortoicetes terminifera (Walker) were wandering in the same direction, fairly close to one another. The large band travelled 45 ft., but the small band only 6 ft. in 1 day. Uvarov (1928) suggested that hoppers in groups were more active than isolated ones, because of an opto-motor reaction : when one hopper moved those near it also moved in such a way as to keep the visual field constant. Field studies have reaffirmed the theoretical importance of opto-motor reactions, although it is clear that other kinds of hopper interaction also play a part (Volkonsky, 1942; Kennedy, 1945, 1951; Clark, 1949).

The optimal conditions for the marching of Locusta migratoria migratorioides in the laboratory have been studied in detail (Ellis, 1951). A group of hoppers will march round and round in a cage, provided there is a single source of light overhead. It is easy to recognize a marching hopper, for the body is carried in a characteristic way with the frons perpendicular to the ground, the antennae errect and the whole body kept well clear of the ground. Marching is most vigorous and prolonged at temperatures between 29 and 35° C., being greatly increased by starvation for 3-13 hr. With fresh grass present it requires the combined effects of radiant heat and air currents, factors which are normally present in the field (Fig. 1).

Even with other facts optimal, marching varies with the number of hoppers per cage. As marching is favoured by ‘mild’ physical conditions, it has been suggested that it is a normal activity of group-living hoppers, which takes place whenever it is not suppressed by an incompatible activity (Ellis, 1951). A similar conclusion was reached by Kennedy (1939, 1951), after studying Schistocerca gregaria in the field. This paper analyses further the problem of hopper numbers and marching, and especially the hopper interactions responsible for the increase in activity with numbers. Only hoppers reared in crowds for many generations were used. Such hoppers are generally said to belong to the phase gregaria, but this term will be avoided, as it has been used to describe locusts with certain anatomical ratios and colours, which are often, but not always, associated with locusts that are living in groups (Key, 1950; Gunn & Hunter-Jones, 1952). The behaviour of hoppers reared in isolation, so-called solitaria, will be dealt with in a later paper.

The experimental animals were 4th instar hoppers of Locusta migratoria migratorioides (R. & F.), bred at the Anti-locust Research Centre, London. The same stock was studied by Hamilton (1936), Norris (1950) and Ellis (1951). The hoppers were reared in a constant temperature room which was maintained at 30-3i° C. during the tests on hopper numbers and densities, and at 27-28° C. during the rest of the study. The hoppers were kept under crowded conditions of 300-400 1st instar nymphs in a cage 43 × 43 by 25 cm. high ; by the 4th instar 200-250 were still alive. Each evening they were fed with grass standing in water, most of which had been eaten by the following morning. During the day a 60 W. lamp was hung centrally above each cage and the hoppers were allowed to march for at least 8 hr. When the room was maintained at 30-31° C., the lamp was water-screened to prevent excessive heating (Fig. 2). Sticks and hay were placed in the corners of the cage to provide moulting perches.

The test hoppers had moulted 2-3 days before use : during the 4th instar, activity is depressed after the 4th day, as moulting time approaches (Ellis, 1951). Individuals were identified by spots of poster paint on the pronotum and abdomen. During the night preceding a test the hoppers were well fed, being starved next morning for various periods before use. When they were needed in a relatively inactive state they were starved for 1 hr., and in an active state for 3 hr. (Fig. 1). During starvation they were placed in jars arranged round an electric lamp. Additional unmarked hoppers were used in some of the tests ; they were the same age as the marked hoppers and were previously well fed: they will be called background hoppers.

During these tests three sizes of cage were used; they were all 25 cm. high and the floor areas were 30 cm. square, 43 cm. square and 60 cm. square. The floors and walls were made of fibre-board and the tops of glass with a triangular gauze lid in one corner. During tests the cages were kept clear of obstructions and food, a 60 W. lamp was hung centrally above each and a water screen (to prevent excessive heating) was placed between the lamp and the cage top. The lamp and water dish were surrounded by a cylindrical black paper shade (Fig. 2). Observations were made through the glass top of the cage, but did not disturb the hoppers inside, provided the lamp was screened and the rest of the room was kept in darkness.

At the beginning of a test the hoppers were transferred from the feeding cages to the experimental cages. The numbers of hoppers per cage were 2,15, 30, 50 and 100, the test at each number being repeated three times. Five hoppers were marked in each cage (in experiments with only two hoppers, both were marked) and during the 6 hr. test they were observed for 10 consecutive min. per $12$ hr. At each reading it was noted whether the marked hoppers were marching or not, and from these readings the percentage of time spent in marching was calculated. Observations which gave the actual speed of the hoppers whilst marching (marching speed) were also made, but will not be considered further. The smaller the cage the lower was the marching speed, suggesting that the hoppers were slowed down by having to travel in small circles. A few tests with one hopper per cage were carried out, continuous observations being made for 1 hr., after they had been starved for 5 hr.

In Fig. 3 the average percentage of time spent in marching for each number of hoppers per cage was plotted against time, there being a separate curve for each cage. It has been shown that the rise in marching activity during tests in foodless cages is related to the emptying of the gut and to a cumulative effect of hopper interactions (Ellis, 1951). On the whole, the curves for each cage were similar with 30 or more hoppers present, but with smaller numbers there was less activity the larger the cage. Thirty per cent of the time was spent in marching after 5 hr. of starvation with two hoppers per cage, after 4 hr. with five hoppers, after $312$ hr. with 15 hoppers, after $212$ hr. with 30 hoppers, after 2 hr. with 50 hoppers and after $112$ hr. with 100 hoppers. That is, the onset of marching was significantly delayed with fewer than 30 hoppers present. With only two and five the maxima were also lower, although it is possible that with longer periods of starvation all maxima would be the same. Ten hoppers were tested by themselves. Two of them did not march during the observation period, whilst three of them marched for over 70 % of the time. The average was 32%, a value very similar to that for two hoppers per cage after the same period of starvation, namely 5 hr.

In order to test the relative importance of the total number of hoppers present and the cage size, the readings for each 6 hr. test were averaged. An analysis of variance on the results showed that numbers significantly affected marching activity (P<0·01), whilst the importance of cage size was doubtful (P=0·02-0·05). Examining the actual means in the light of this analysis, the overall mean for the 60 cm. cage was significantly lower than that for the 30 cm. one. The overall means for the numbers of hoppers per cage differed significantly for each increase up to 50: there was no significant difference when the results for 50 and 100 hoppers were compared. A similar conclusion is to be drawn from Fig. 4 in which the percentage of time spent marching was plotted against (A), the log₁₀ of the total number of hoppers present and (B), the log₁₀ of the number of hoppers per square metre of cage floor, called area-density for short. Logarithms were used in order to foreshorten the curves at the higher numbers and area-densities, where activity is unaffected by numbers. With a given number of hoppers present, marching was greater in the small cages (Fig. 4A), but with a given area-density marching was greater in the large cages (Fig. 4B). This again shows that numbers and areadensity were important, although with more than 50 hoppers per cage marching may have depended on numbers rather than area-density.

These tests clearly showed that vigorous marching was a group activity, which was greater in large and closely packed hoppers groups than in small and more scattered ones. The rest of this paper deals with the hopper interactions which lead to this increased marching activity.

When marching is well established the antennae are kept still and seem rarely to be used, but at the beginning of marching, when hoppers spend some time wandering about at random, the antennae are continually used to examine the floor and other hoppers. A resting hopper examined by an active one, kicks with its hindlegs if touched on the abdomen, but twirls its antennae in a face-to-face encounter. These reactions generally result in the active hopper moving away.

The marching activity of antenna-less, 4th-instar hoppers was compared with that of normal individuals. The experimental hoppers were marked in the usual way and the tests were carried out in cages 43 × 43 by 25 cm. high, made of fibreboard, but with a glass top. A 60 W. lamp, water-screen and cylindrical shade were placed centrally above the cage (Fig. 2). The room temperature was maintained at 30-31° C.

The 16 marked hoppers and 40 background hoppers were well fed overnight and were placed in the test cage next morning. During the next 6 hr. marching by the marked hoppers was noted once per min., for 10 consecutive min. per $12$ hr. ; from these readings the percentage of time spent marching was calculated. The complete test was repeated 3 times.

The results for the eight antenna-less hoppers were averaged for each day and compared with the corresponding averages for the eight control hoppers. These averages are shown in Fig. 5, in which the percentage of time spent marching was plotted against time. The two curves represent the means for the three tests. Amputation of antennae had no marked effect on marching, although the graph suggests that the antenna-less hoppers were a little more active during the first 3 hr. of test, than the normal hoppers.

The marching behaviour of 4th-instar hoppers which had had their eyes painted over with poster paint for 20 hr. was compared with that of normal animals. Both kinds of hopper were marked and starved for 5 hr. before being dropped into a cage containing a small marching band of hoppers. A water-screened lamp was placed centrally above the cage (Fig. 2) and the air temperature was maintained at 30-31° C. The test hoppers were observed once per minute during the hour after dropping in. Activity was classified as marching, pottering (random movement) or resting, and the percentage of time spent in each of these activities was calculated from the 1 min. readings.

At the end of the tests the blinded hoppers were examined with a hand-lens, and any with defective eye coverings were excluded from the final results : four out of 32 hoppers had to be discarded. The complete experiment was repeated on four different days.

Three of the discarded hoppers performed circling movements for long periods, turning towards the eye which was partially uncovered. Total blinding did not prevent hoppers feeding. On one occasion a blinded hopper rapidly joined with others in devouring a newly moulted fellow and at the end of the tests, when grass was put into the cage, the blinded hoppers were soon feeding with the rest. Presumably, the blinded hoppers moved towards the food in response to olfactory or auditory (sound of other hoppers chewing) stimuli (cf. Volkonsky, 1942).

Twenty-four normal hoppers were tested, and all of them spent some time in marching, the individual results varying between 42 and 83%. Only 58% of the blinded hoppers marched, and for only 8-20% of the time. Generally they appeared to be kept moving in the direction of the general stream by repeated impacts from other hoppers, but four blinded individuals did march for short periods without being pushed. The normal hoppers spent 10-28% of the time pottering, whilst the blinded hoppers spent 2-20% of the time in this activity. The 42% blinded hoppers that did not march went to the cage sides; half of them remained still throughout the test period, whilst the other half averaged three changes in position and pottered for up to 10% of the time.

Although it is tempting to suppose that these experiments suggested the importance of optical stimuli between hoppers in increasing marching, the eyes may nevertheless act as ‘stimulationsorgane’, so that stimuli received via them lead to a general increase in locomotor activity (Wolsky, 1933; Chauvin, 1947).

The parts played by optical interactions and physical contact between hoppers (called mechanical interactions for short) in increasing marching activity were tested in a series of five experimental arrangements that eliminated one factor at a time. Three types of experimental cage were used, each containing a circular, ring-like test gallery, 8 cm. high, 8 cm. across, with an inner diameter of 28 cm. and an outer one of 44 cm. The floors were made of fibre-board and were marked into eight equal divisions by pencil marks. The tops and sides were made of celluloid.

• (1) The single gallery cage consisted of a test gallery with the outer walls backed by white cardboard (Fig. 6A).

• (2) The treble gallery cage consisted of a test gallery surrounded by an outer gallery 5 cm. across and an inner gallery also 5 cm. across (Fig. 6B). The outer walls of this cage were backed with white cardboard.

• (3) In the moving pattern cage the test gallery was suspended over metal spokes which carried an outer and an inner cylinder of white cardboard (Fig. 6C). The spokes were soldered to a bicycle wheel hub, mounted upright on a heavy wooden base. The cardboard cylinders carried one of two patterns, which were turned by hand, in an anti-clockwise direction, at about 3 r.p.m.

• (a) Vertical stripes. There were 64 alternately black and white stripes, 2·5 cm. wide, on the outer cylinder and 64 stripes, 1 cm. wide, on the inner cylinder.

• (b) Dead hoppers. Thirty-two dead, 4th-instar hoppers were stuck 2·5 cm. apart on the outer cylinder and 18 on the inner cylinder, facing in an anti-clockwise direction. The dead hoppers were prepared as follows. After killing, the soft exoskeleton in front of the thorax was pierced and the body fluids squeezed out. The hoppers were then filled with warm wax blown from a glass tube. When dry, these dead hoppers retained their rounded form.

During the tests, a 100 W. lamp, surrounded by a white cardboard shade, 50 cm. high and 60 cm. in diameter, was placed 70 cm. above the floor of each cage.

The five experimental arrangements were :

• (1) Hopper contacts. The single gallery cage contained 50 marching background hoppers.

• (2) Gallery hoppers. There were 20 marching background hoppers in the inner gallery and 60 in the outer gallery of the treble gallery cage.

• (3) Dead hoppers. The dead hoppers stuck to cardboard were moved round outside the moving pattern cage.

• (4) Vertical stripes. The vertical stripe pattern was moved round outside the moving pattern cage.

• (5) The controls were of three kinds. The single or treble gallery cages empty, or the moving pattern cage with plain white cardboard cylinders outside.

The differences between pairs of arrangements were as follows. In hopper contacts the test individual gave and received both mechanical and optical stimuli from the marching background hoppers in the cage, whereas in gallery hoppers there were only optical reactions between the test individual and the hoppers in the outer and inner galleries. Any differences in the activity of the test hopper when placed in these two cages was therefore due to the mechanical interactions between this individual and the background hoppers. In a similar way the other arrangements analysed the optical reactions between hoppers; gallery and dead hoppers the importance of live hopper movements ; and dead hoppers and vertical stripes the importance of hopper shape plus colour. The controls gave the activity of isolated test hoppers not subjected to any special stimuli. The results for all arrangements must be compared with the controls.

The 4th instar experimental hoppers were tested individually, and during the 5 min. observation period the divisions of the gallery floor travelled clockwise and anti-clockwise were counted. The number of seconds that the hopper spent in activity was also recorded, but was difficult to measure accurately. Most of the walking was marching, although some random movement (pottering) took place : no hopper pottered for more than two divisions of the cage floor per reading. Marching hoppers moved fairly continuously and, especially in the dead hopper and gallery hopper arrangements, tended to keep to a straight path as long as possible (Fig. 6 A). In the vertical stripes arrangement many hoppers kept close to one side of the gallery and often scrambled up the wall. With gallery, dead hoppers and vertical stripes arrangements, test hoppers frequently looked squarely out of the gallery and pointed their antennae towards the objects outside.

The data were analysed in six ways. First, the percentage of hoppers that moved during the observation period was calculated and called the percentage active. Unfortunately, the individual results were not normally distributed, neither did they follow any other distribution that could be analysed statistically (Table 1), so the other five analyses were based on AVERAGES for the ACTIVE hoppers only. As each division of the cage floor was approximately 14 cm. long, the distance in cm. per min. travelled clockwise (C), anti-clockwise (A) and (A + C) were calculated separately, by multiplying the divisions of the cage floor travelled per 5 min. reading by 14 and dividing by 5. This measure of activity was called progress speed, and a change in its value could be due, either to a change in the number of seconds per reading that the hopper was active (the seconds active) or to a change in the rate at which the hopper moved whilst actually marching (the marching speed). The seconds active were obtained by dividing by 5 the total number of seconds active per reading. The marching speed was calculated by adding the progress speeds A and C, irrespective of direction, multiplying the product by 60 and dividing by the seconds active, to obtain the result in cm. per min.

Each test hopper was placed in the apparatus for 10 min., observations being made during the second 5 min. During the first 5 min. the hoppers behaved abnormally because of recent handling; that is, the percentage active was higher and the progress speeds and seconds active were generally lower than later on. This is illustrated in Fig. 7, showing the results for tests in which hoppers were observed for 30 min., after being placed in the hopper contact or vertical stripes arrangements.

In order to present the five experimental arrangements to the same individuals in one day, the hoppers were tested after 1 hr. of starvation, fed for 20 min., starved for 1 hr., re-tested and so on. As the test period lasted for 10 min., seven or eight individuals were used in turn. The order in which the five arrangements were presented to the hoppers was altered with each day’s test, for even when using the same arrangement {vertical stripes) throughout the day, the progress speeds and seconds active increased with each successive test (Table 2).

When comparing the results for the various arrangements it would be unwise to treat any differences as significant, unless they are greater than those in Table 2: for example, 16 points for the percentage active and 10 points for the seconds active.

The percentages of hoppers that reacted during the test period were in three groups (Fig. 8), which differed by more than 16 points from one another. Only 30% of the hoppers reacted under control conditions, whereas over 80% did so with hopper contacts. Vertical stripes, dead and gallery hoppers were intermediate in effect and approximately 50% of the hoppers reacted to them. This means that, even with no special stimuli, some hoppers that have been starved for 1 hr. moved ; but more did so in response to a moving background and nearly all of them moved when in actual physical contact with other marching hoppers.

The amount of activity shown by the active hoppers was measured by the progress speeds A + C, the results forming three groups, which differed from those for the percentages active. The two extremes, which can be considered to differ significantly, were the controls and the dead hoppers plus gallery hoppers plus hopper contacts. The results for the vertical stripes were intermediate between the two extreme groups (Fig. 8). Thus, progress speeds were increased when the hoppers were subjected to a moving background of stripes, but the maximal amount of activity was only reached in response to moving hopper bodies, alive or dead. Physical contact between hoppers did not increase progress speeds further. It has already been pointed out that the progress speed depended on the seconds active and the marching speed. The results for the seconds active were similar to those for the progress speed A + C, but the marching speeds did not vary a great deal as between the various arrangements, although the controls gave the lowest and the dead hoppers the highest results. It therefore appears that the differences in the progress speeds A + C depend on the amount of time per reading that the hoppers spent in activity, rather than the rate at which they moved when actually marching.

Frequency distribution tables of progress speed A + C and seconds active (Table 1) show that the average increases in these measures depend on a decrease in the percentage of hoppers moving short distances and an increase in the percentage moving long distances. The results in these tables again suggest the three groups of the controls, vertical stripes, and gallery hoppers plus dead hoppers plus hopper contacts.

The average progress speeds travelled in both directions are shown in Table 3. Variations in the progress speeds A were similar to those for the progress speeds A + C (Fig. 8). Progress speeds C showed no clear trend, but were interesting when compared with the progress speeds A. The ratios between these two measures (Table 3) suggest that moving test hoppers had little preference for direction under control conditions, but that they showed a definite preference for the direction in which visible objects moved. Not only did a moving background increase marching activity, but it caused marching to be in a certain direction. The relative increase in the counter direction walking with hopper contacts is not easy to explain.

Marching hoppers in a cage are clearly audible to the human ear, most of the noise being due to the movement of the hoppers’ feet. In common with other insects, locusts respond to a wide range of sounds (Pumphrey, 1940), and Kennedy (1951) suggested that the noise made by moving locusts helps to increase the total activity of the group. Although the last experiment showed that dead hoppers, artificially moved, stimulated marching to the same degree as live hoppers marching nearby, it is possible that hoppers respond to ground vibrations received via the tarsi from other individuals (Chauvin, 1947), when they cannot actually see one another.

The marching activity of isolated hoppers able to see marching individuals nearby was compared with their activity when the nearby hoppers were invisible. The circular, ring-like cage, divided into three concentric galleries was used (Fig. 6B). Three arrangements were compared, the controls, gallery hoppers and a new arrangement, screened hoppers, in which strips of white cardboard were placed outside the transparent walls of the centre gallery and 20 background hoppers were allowed to march in the inner gallery and 60 in the outer one. A test hopper in the centre gallery could hear and receive ground vibrations from the marching hoppers outside, but was unable to see them.

The test hoppers were starved for $112$ hr., tested in one of the arrangements for 10 min., fed for at least 20 min., starved for $112$ hr., re-tested in another arrangement and so on. Each hopper was presented with the three arrangements on one day and six to eight individuals were used in turn. The hoppers were allowed to settle down during the first 5 min. in the apparatus, and during the second 5 min. the divisions of the gallery floor travelled anti-clockwise and clockwise were counted. In 3 days of experiments 23 hoppers were tested.

The results were analysed in three ways (Fig. 9). The percentage of hoppers that moved during the observation period was lowest in the controls, highest when the marching hoppers were visible and intermediate between these two with screened hoppers. During these tests, the background hoppers in the inner and outer galleries marched in an anti-clockwise direction. The progress speeds were averaged for the ACTIVE hoppers only ; speeds in the clockwise direction were similar for the three arrangements, whilst those in the anti-clockwise direction were far higher with gallery hoppers than with the other two arrangements. In the screened hopper arrangement the test hoppers spent a great deal of the time scrambling up the wall of the inner gallery as if they were trying to go towards the hoppers outside.

The proportion of active isolated hoppers was increased in the presence of other hoppers nearby which were invisible, but sustained marching by the isolated hopper depended on the others being visible to it.

Under field conditions, marching is characteristic of hopper bands. The laboratory experiments clearly illustrated an increase in marching activity with increasing numbers of hoppers per cage. Although isolated hoppers marched, averages reached only 32% after 5 hr. of starvation and behaviour varied enormously from individual to individual. This is to be compared with hopper groups of 30 or more, which marched for over 60% of the time after 3 hr. of starvation. The physical conditions which favour marching are not extreme (Kennedy, 1939, 1951; Ellis, 1951) and appear to provide conditions under which marching can express itself. The influence of other hoppers, however, is so great that this factor might well be called a stimulator of marching: a similar conclusion was reached by Clark (1949). As hoppers have a profound effect on each other’s activity, those factors which bring and hold the members of a band together are of particular importance. The present experiments illustrated a tendency for hoppers to keep together, but also a tendency for them to spread over the space available, so that marching depended on both cage size and the number of hoppers present. It is not easy to apply these results to field conditions, where there must be a balance between factors causing the disruption and those causing the cohesion of bands. This problem has not been studied in detail, but disrupting forces include heavy rain, cloudy weather and dense vegetation (Kennedy, 1939; Clark, 1949), whilst factors favouring band cohesion include the attraction of hoppers to the same physical environment (Kennedy, 1939), the tendency for hoppers to move parallel with one another (Table 3 and Kennedy, 1951) and a positive attraction of individual hoppers towards groups of their fellows (Clark, 1949).

The experiments on hopper numbers and density illustrated the dependence of marching on both internal and external conditions and the way in which they balanced each other. In the past, a variety of conditions have been shown to play a part in the full expression of marching; e.g. the moulting cycle, the degree of crowding during rearing, the period of starvation, temperature, radiant heat, air currents and other hoppers (Ellis, 1951). In the present tests the two varying factors were the length of starvation and the number of other hoppers. Activity increases with a decrease in the amount of food in the gut (Ellis, 1951) and the ^ramount of stimulation provided by other hoppers presumably depends on the number of contacts per unit of time, which in its turn varies with the number of hoppers in the cage. Reference to Fig. 3 shows that the time taken for 30% of the hoppers to be marching varied from 5 hr. of starvation with only two hoppers per cage to $112$ hr. of starvation with 100 hoppers per cage. For the same amount of marching, a hopper in a small group (low external stimulation) had to reach a higher level of general activity (i.e. the threshold for sensory stimulation was lowered by starvation for a longer period) than one in a large group (high external stimulation). Presumably, in the field, any one of the factors listed above may limit marching, depending upon the conditions at the time.

The analysis of the fully developed marching behaviour pattern illustrated the importance of optical interactions in increasing hopper activity and supported the theory put forward by Uvarov (1928). Locust hoppers, in common with many other insects, carry out compensating movements when the background is moved (Loeb, 1918). Hoppers appear to experience ‘discomfort’ when images move across the eye in an anterior-posterior direction, or rapidly in a posterior-anterior direction. The resulting compensating movements cause the hoppers to move in the same direction as the background (Kennedy, 1951); Table 3 illustrates this tendency.

In the laboratory, a moving background increased the proportion of active hoppers and the distance travelled per minute : the progress speed. Vertical stripes and live hoppers marching produced the same proportion of reactions, but the progress speeds of the reacting hoppers were higher with moving hopper bodies than with vertical stripes. Hoppers began to react to the stripes, but the movement was not sustained. Live hoppers marching and artificially moved dead hoppers gave similar results, suggesting that the jerky movement characteristic of marching hoppers did not play an important part in the opto-motor reaction. The features of hopper appearance (such as shape, size and colour) which bring about the maximal progress speeds will be discussed in a later paper : hopper bodies may be more effective than vertical stripes only because their longer outline stimulates a greater number of adjacent ommatidia at any one time (Zerrahn, 1934).

Mechanical interactions between hoppers were necessary for the maximal amount of marching by the maximal number of hoppers in a band. Mechanical interactions increased the proportion of active individuals very considerably, but did not increase the progress speeds of the active hoppers above those obtained with optical stimuli alone. This may indicate that the opto-motor reaction elicits the maximal amount of marching in some individuals, so that further stimulation has no apparent effect: mechanical interactions would then elicit responses in other individuals which will not react to optical stimuli alone. However, mechanical interactions not only caused resting hoppers to move, but they also slowed down active individuals in one of two ways. First, marching hoppers frequently collide; this is always followed by a few seconds pause by the hoppers concerned. Secondly, if a marching hopper comes upon a resting one, it frequently pauses to examine the resting individual. The few blinded hoppers that marched suggest that, on rare occasions, mechanical interactions alone may produce sustained marching; but most of the experimental results showed that mechanical interactions raised the general activity of the hoppers, while sustained movement in one direction depended on an opto-motor reaction to a moving background.

The other interactions that were tested did not play an important part in marching. Air and ground vibrations from marching hoppers tended to increase the general activity of isolated individuals not subjected to a moving background and may possibly play a part in the movement of bands at night, or under conditions where the hoppers are momentarily obscured from one another.

The laboratory experiments were designed to test increases in marching activity due to optical and mechanical reactions between hoppers. Kennedy (1939, 1951) suggested that such reactions would also keep bands together. It is difficult to agree with this view without appropriate laboratory tests having been made. An examination of Table 3 shows that some hoppers moved against the general stream. In the field, such behaviour by individuals at the edges of the band would lead to a slow dispersal of the hoppers during marching. There are no field observations on this point, although it seems to be assumed that bands do not disperse during marching. If this is the case, then it is reasonable to look for specific reactions that keep bands together, such as a movement towards, instead of parallel with, hopper bodies considerably smaller (farther away) than the hopper itself.

This study has interesting connexions with the feeding of hoppers in the field. Dense populations of locusts could live either fairly evenly dispersed over a wide area, or concentrated into bands so that their distribution is discontinuous. If they are to avoid starvation, dense populations of hoppers must move about actively. Concentration into bands has the advantage that it increases hopper activity and the distance travelled per day. Hoppers of Locusta migratoria march under normal conditions; but if food becomes short, then after a few hours of starvation very vigorous marching takes place, which should increase the chances of the hoppers coming upon a new food supply. It seems that starvation does not normally play an important part in the marching of this species, but in Chortoicetes terminifera (which in many ways is intermediate between grasshoppers and locusts like Locusta migratoria) the hoppers generally march after they have eaten all the food in one area (Clark, 1949). In more typical locust species marching occurs before the food in one area has been eaten, probably because their hopper interactions result in greater activity than those of Chortoicetes terminifera. Comparative studies of species which are intermediate between typical locusts and grasshoppers may indicate how the mass movements of locusts have evolved.

I have to thank the Anti-locust Research Centre for grants that enabled me to carry out this research; Prof. D. M. S. Watson and Prof. P. B. Medawar for giving me facilities to work in the Zoology Department of University College, London; Dr B. P. Uvarov and Dr D. L. Gunn for helpful discussions; Mr Hunter-Jones and others at the Anti-locust Centre for the supply of animals and Mr Redpath of University College for making the moving background cage.