The Orientation of Ants: II. Orientation to Light, Gravity and Polarized Light

In a previous paper experiments with the ant Myrmica ruginodis were described (Vowles, 1954): in these experiments light and gravity, and light and polarized light, were interchanged as orientatory stimuli. Under such conditions successive orientations were correlated in a rather complex way. In the light/gravity substitutions the orientation angle to the second stimulus (θ₂°) was related to the orientation angle for the first stimulus (θ₁°) in one of four ways expressed by the equations:

The causes underlying the occurrence of these four relationships will be examined in this paper.

It has been suggested by Crozier and others that, for many animals, orientation to gravity on an inclined plane is simply the result of the animal turning upwards until it is in stable equilibrium; this being attained when the centre of gravity is within the triangle of support. Work with the beetle Tetraopes (Crozier & Stier, 1929) supported this idea, and suggested that the weight of the abdomen was the force most important in disturbing equilibrium. Barnes (1929, 1930 a, b) understood his own results with ants to confirm Crozier and Stier’s hypothesis, but recorded also two observations of which he did not recognize the significance:

• I. Ants which had their antennae amputated could not maintain any particular orientation when walking on an inclined plane.

• II. When walking up an inclined plane ants often oscillated successively left and right of the vertical plane, thus making their track a zigzag placed symmetrically either side of the actual upward direction.

The significance of these observations will become clear later. Similar observations were made with Myrmica laevinodis during the present study.

In Crozier’s sense orientation to gravity is a simple geotaxis, in which the animal does not need to perceive the direction in which gravity acts, but merely to turn until the strain of maintaining equilibrium is at a minimum. The maintenance by an ant or a bee of a constant orientation (other than 0° or 180°) on a vertical surface is not, however, of this nature. An ant orientating in this way is usually in unstable equilibrium, for the tip of the abdomen is not trailed on the ground while the insect is walking, which Barnes suggests, and the centre of gravity is always posterior to the support given by the legs. An ant, orientating to gravity, is performing an activity similar to the ‘light compass reaction’. It is proposed to call this activity a ‘force compass reaction’, which may be defined as ‘the orientated locomotion of an animal during which its anterio-posterior axis makes any constant angle with the line of action of a physical force; the maintenance of which orientation does not depend on equal bilateral stimulation, or on keeping a stable equilibrium’. Such an orientation is a ‘menotaxis’ in Kuhn’s (1919) terminology.

This type of orientation raises the problem of the sensory mechanisms by which an ant can perceive the direction in which gravity acts. If an ant were sensitive to the magnitude of the rotational forces, due to gravity, which act upon it, this would provide a basis for geo-perception, for these forces vary with the sine of the angle of deviation from the vertical. If the ant perceived only the magnitude and not the direction of these rotational forces, then a confusion might arise between tracks symmetrically placed on either side of the vertical, when the sines of the angles of orientation are equal. These angles are θ°, 360–θ°, 180–θ°, and 180+ 0°. It is a confusion between precisely these angles that might be expected from the results of the light/gravity substitutions.

It was therefore decided to investigate the orientation of ants to gravity, and to test particularly the hypothesis that confusion between various orientations does occur.

The apparatus consisted of a simple, circular turntable, its surface in the vertical plane, which could be rotated about its centre in this plane. The turntable was in. in diameter, and was placed against a circular scale, which indicated the angle through which it had been turned. The surface of the table was covered with white Bristol board, ruled lightly into 1 in. squares; it gave a good foothold to the ants.

It was found that ants showed little tendency to remain still when on a vertical surface. They would walk in a straight line after very slight stimulation—little was required of the experimenter beyond turning an ant back from the table’s edge. A Perspex screen was therefore placed 3 in. away from the table, and parallel to it: the ants’ tracks were plotted on this in various shades of ‘lipstick’. The red colour of this was convenient, as it disappeared in red light and did not confuse the observer.

On one comer of the Perspex sheet was placed a card on which had been drawn a circle with diameters marked in different colours every ten degrees. By glancing from the ant to this device the observer could make a rough estimate of the actual orientation of the ant, and obtain some idea of the angle through which the table must be turned.

The apparatus was set up in a dark-room in a dim red light. An ant was placed on the turntable and allowed to wander for 2 min. in order to become accustomed to the conditions. The ant was then gently stimulated and set off in a straight line. Its track was plotted. When it had travelled about 4 cm. the table was smoothly turned into a new position and the track of the ant again plotted. When possible the turning was done when the ant was near the centre of the table. After one such ‘run ‘had been completed the plotted track was given a reference number, and the angle through which the table had been turned recorded. The process was then repeated using a different shade of lipstick for the new track. About four recordings were made with each ant, which was selected at random from three colonies. The angles of orientation for each part of the track were then recorded, being measured clockwise from the vertically downward direction.

Three series of experiments were done in order to test the possible equivalence of various orientations.

• I. When an ant was orientating at θ° it was turned to face in a direction at 180-θmain regression along the lin°.

• II. When an ant was orientating at θ° it was turned to face in a direction at 180 + θ°.

• III. When an ant was orientating at θ° it was turned to face in a direction at 360-θ°.

The turning was always done the shortest way. With practice the experimenter became fairly accurate in turning the table through the desired angle. Of all the attempts so made about half were accurate to within 15°.

The species of ant used in this and subsequent experiments was M. laevinodis.

It was found that an ant would walk in almost any direction relative to the vertical. There was, however, a tendency to avoid walking at 90° or 270° or angles approximating to these, in which position the strain on the ant’s legs on one side is presumably maximal.

The orientation angles to gravity before and after turning, were compared for each of the three series. It was found that Series I and II did not differ significantly from each other. In the lower graph of Fig. 1 the orientation angles before and after turning are plotted against each other for the experiments of Series I and II. In the upper graph are plotted the same angles for the experiments of Series III, in which the ant was turned from θ° to 360—θ°, i.e. turned through 20°. The lower graph will in future be called the control graph, and used for comparison with the upper graph.

Consideration of the control graph shows three effects:

• I. A main regression along the line θ₁= θ₂.

• II. A few points lying about the line 360—θ₁ = θ_S.

• III. A broad band of points lying about the value 180° on the ordinate. When ants were disturbed they often turned and walked in a general upward direction: it is this behaviour which gives rise to this horizontal, broad band of points.

As all these relationships occur on the same graph it was thought that any statistical estimate of the correlation would have little real significance. The graph may be interpreted qualitatively as showing that:

• I. In most cases an ant returns to its original orientation after the table has been rotated.

• II. Occasionally an ant turns to a new orientation, expressed by 360—θ₁ = θ₂, after rotation of the table.

• III. There is a tendency for ants to walk upwards after being disturbed by movement of the table.

The main importance of this control graph lies in providing a comparison with the upper graph. Consideration of the upper graph shows the points to be scattered about the two arms of a cross, the equations of which could be expressed by:

and

The first equation would be the result of the ants returning to their original orientation after the turning of the table; and the second equation the result of the ants turning very little after the table had been rotated.

There are no points at the centre of the graph because when an ant was orientating at about 180⁰ the table did not have to be turned, and the ant was not therefore disturbed. It was thought that such results, where the ant walks straight on, would not contribute to the analysis, and they were omitted. In addition to the points lying along the arms of the cross there are others scattered apparently at random, and others showing that an ant turned and walked upwards after the rotation of the table.

No statistical analysis of the graph has been attempted, as it is considered that the relationships are fairly obvious.

It was expected that the points falling about the line θ₁= θ₂ would come from those results when the table had not been turned through the required angle (2 · θ₁), while those on the line 360—θ₁ = θ₂would come after successfully turning the table. This was found not to be so, as is shown in Table 1 below.

The first orientations appear not to influence which of the two possible new orientations shall be chosen, as can be seen from the graph itself.

An attempt was made to calculate the accuracy of the gravity orientation from the control experiments. A graph was plotted of angle through which the ant turned/angle table turned. Again a main regression was found corresponding to the line θ₁= θ₂ on the original control graph: but the other points were now scattered widely over the whole area of the graph. Those points obviously not falling along the main regression fine were then omitted from the calculations. The calculated regression was

The standard error of 17·69 was ± 9·29, and this constant therefore does not differ significantly from o.

The standard error of 0·8545 was ± 0·0649, and this figure is therefore significantly less than 1. This implies that the ants turn through slightly less than the angle through which the table was turned.

90% of the points fell within + 33⁰ of the regression line. This may be taken as an indication of the possible accuracy of orientation to gravity.

The experiments show that if an ant is orientating to gravity, and is turned away from its original heading, it will usually return and again take up its original orientation. However, in the experiments, when an ant had been orientating at θ° and was turned toward or past a position in which it would have been heading at 360—θ°, it only resumed its original orientation on about half these occasions; on the other half it took up a new orientation at 360—θ°. There was also a tendency for ants to do this in control experiments. This agrees with Barnes’s observations that ants often oscillate, giving a zigzag path about the vertical. This was also seen in M. laevinodis.

It seems therefore that ants somehow confuse orientations lying symmetrically on either side of the vertical. This could be due to either central or sensory factors or both. It is possible, however, to imagine a situation in which such confusion would be the direct result of the physical factors involved in georeception. Suppose, for example, that the ant was sensitive to the distorting force due to gravity upon some part of its body. These forces will vary with the orientation of that part of the body relative to the vertical: the rotational force will be maximal when the axis is at 90° to the vertical, and minimal when it is parallel. The size of the rotational force is proportional to the sine of the angle between the axis and the vertical. Therefore for any one position in each quadrant there is a corresponding position in each other quadrant where the rotational forces are the same. If the ant could distinguish the magnitude but not the direction of these rotational forces it would confuse four orientations placed symmetrically about the vertical. However, the distorting force due to gravity also has a longitudinal component acting along the axis of the part of the body involved: the direction of this component is directly opposite when the axis points in a general upwards, or a general downwards direction. Therefore if the ant was sensitive to the magnitude but not the direction of the rotational force, and to the direction but not the magnitude of the longitudinal force it would confuse two orientations placed symmetrically left and right of the vertical.

If this hypothesis is taken as a working basis for experiment one must consider various joints of the body across which strain could be measured. Those considered here are

1. The abdomen to the thorax.

2. The thorax to the head.

3. The limbs to the thorax.

4. The scape of the antenna upon the head.

5. The funiculus of the antenna upon the scape.

The experiments described below were designed to locate the georeceptor within the ant.

Two standard methods used in such experiments—extirpation of parts, and splinting of joints—could not be used here. All the parts of the ant to be considered have important functions other than geosensory, and both extirpation and splinting caused general behavioural disturbance, which rendered the operated individual unsuitable for experiment. Moreover, it is often unwise to make inferences about the normal function of an organ based solely on observations of an animal’s behaviour after it has been deprived of that organ.

The method used here was to vary the actual rotational forces upon different parts of the ant. This was done by cementing a small particle of soft iron to the part of the ant to be studied, and while the insect was walking on a vertical surface subjecting it to a magnetic field. Care was taken to use a minimum of shellac cement, particularly on the antennae.

The apparatus consisted of a flat board, 25 cm. wide, fixed in a vertical position. The board was painted dull black, and on its surface a filter paper, lightly pencilled into 2 in. squares, was glued, to provide a good foothold for the ants and a good background for the observer: being circular and of large diameter (23 cm.) the filter paper had no sharp comers which might have acted as landmarks for the ant. At either end of the transverse axis of the board was placed a large flat-wound solenoid, mounted on brass formers, 8 in. in diameter and in. wide. The solenoids were placed with their planes at 90° to the surface of the board.

Each solenoid had 100 turns of wire, and a resistance of 4 Ω. They were connected in parallel with a 4 V. accumulator. The current passing through each of them was i A. This arrangement gives a fairly uniform magnetic field of strength approximately 3 G. The fines of force lay horizontally across the surface of the board. The effect of the magnetic field was to produce a rotational couple tending to turn an iron rod into the direction of the field. This couple acted on the same direction, clockwise or anticlockwise, when the axis of the iron rod lay in diagonally opposite quadrants; and in the opposite direction in adjacent quadrants. The apparatus was set up in a dark-room, in a dim red light, with the earth’s magnetic field acting in a plane at 90° to the surface of the board: the earth’s field therefore could not influence orientation in the plane of the board.

An ant prepared for the experiment was placed on the filter paper, and was allowed to wander at random for 2 min. The escape reaction was then released, as in the turntable experiments. When an ant had travelled for a few centimetres the magnetic field was switched on and the behaviour of the ant observed. The procedure was then repeated with the field on initially, and switched off while the ant was running. About six such observations were made with each ant.

It was observed that ants with the iron filings cemented to them did not behave differently from normal ants when on the vertical surface, when the magnetic field was off or when it was on all the time. It seems that ants can orientate satisfactorily under both conditions. The results of the experiments are summarized in Table 2.

The fact that some sort of response was shown by the ant in all cases, except when filings were on the thorax, suggests that the magnetic force exerted was sufficiently strong to be perceived by the ant, even when it had no effect on orientation. An attempt was made to calculate this force and to compare it with the forces due to gravity: a number of assumptions had to be made about the dimensions of the filings used, which were not uniform, the inductance of the iron, which was not known, the size of the ants, which varied, and the exact position in which the filing was attached to the ant; such assumptions of course decrease the accuracy of the estimation, but it may be said that the maximum magnetic couple is of the order of 10 · the rotational force due to gravity upon one antenna. It seemed possible therefore that this force was too small, compared with the weight of the head and abdomen, to influence orientations depending on these members. The fact that the magnetic force caused other behavioural effects, such as swerving, biting and stopping, argues against this possibility, but does not eliminate it. A few experiments were therefore done using a strong bar magnet: an ant with filings on either its head or abdomen was allowed to run on a vertical surface, and then the magnet was suddenly placed diagonally behind it; the force exerted was strong enough to pull the head or abdomen sideways; this usually caused the ant to stop, but occasionally it would continue to run, and in such cases preserved its old orientation: this supports the hypothesis that the head and the abdomen are not concerned with orientation to gravity in ants.

The results can be interpreted as follows: the georeceptor is situated somewhere between the funiculus and the scape; the perception of the direction in which gravity acts, relative to the ant, depends on the rotational force exerted on the funiculus; when an iron filing is present on the funiculus, and a magnetic field is operating on the filing, the stimulus for orientation is the sum of the rotational forces due to gravity and the magnetic couple. Thus, when a magnetic field is switched on or off an ant turns until in a position where the rotational forces are restored to their original value. When an iron filing is present on the scape, change of magnetic field may cause actual movement of the antenna, leading to a temporary change in the rotational forces on the funiculus, which persists until the ant restores its antenna to its normal position; such a change would lead to temporary disorientation and swerving as seen in the experiments.

When an ant is running freely on a vertical surface it holds its antennae in a fixed position. The scape is held diagonally forward and outward, at about 30° to the anterio-posterior axis of the ant; the funiculus is held more nearly parallel to this axis, but is turned slightly outward from it, particularly at the club-shaped tip. The relative positions of the antennae to the head appear the same at all orientations: the rotational forces upon the funiculus are therefore not equal for two orientations lying symmetrically either side of the vertical, for the two symmetrical positions for the ant are not symmetrical for the funiculus. The confusion between such orientations cannot, therefore, be due to equal rotational forces upon the funiculus, and needs another explanation.

It seemed possible that the explanation might lie in the interaction of the two antennae. In the experiments so far described an iron filing was placed on both antennae. It was not known, however, what would happen if an iron filing was placed on only one of the antennae, when the two antennae would be unequally influenced by a change in the magnetic field. Some experiments were therefore done with ants which had an iron filing fastened to only one antenna. The experimental procedure was then the same as in the previous experiments.

It was found that when the magnetic field was switched on or off an ant was usually disturbed in some way, sometimes it smoothly changed direction, and sometimes it swerved but then returned to its original orientation. The results are summarized in Table 3. If, when the ant smoothly changed direction, the two parts of the track were symmetrically placed either side of the vertical it was assumed that the ant had turned spontaneously making part of a normal zigzag path, and in the Table such behaviour is counted as if the ant had not turned.

If the magnetic field was changed more than once during a single run it was seen that an ant sometimes responded to only one of these changes by turning; whether the operative change was the first or the second could not be predicted, it varied both with different ants and with the same ant on different occasions. No significant difference was observed between turns caused by the presence of the iron filing on one antenna, and those when both antennae were affected.

These results suggest that the georeceptors function independently in the two antennae, and are used singly, the receptor on one side being dominant at any one time; if this is so it is a very unusual condition. A change from one antenna to the other may possibly occur during an orientated run, causing a change from the original to its equivalent direction. When an ant is orientating at θ° the magnitude of the rotational force due to gravity on one funiculus is equal to similar forces on the other funiculus when the ant is orientating at 360—θ°. If one assumes that equal forces on the two antennae produce the same effect on the taxis mechanism, then the equivalence of two such orientations could be due to the use of one antenna for one orientation and the other for its equivalent; the zigzag path would then be the result of successive changes from one to the other organ. The turntable experiments suggest that the change can be initiated by swinging the ant toward or past the vertical.

The experiments using iron filings located the georeceptor between the scape and the funiculus of the antennae. The only organ present in this region, which could fulfil the functions of a georeceptor, is Johnston’s organ, which is situated in the pedicellus. This organ, which consists of scolopidia or modified chordotonal organs, was first described by Johnston (1855) in Culex. Child (1894a, b) showed it to be present in Hymenoptera in a poorly developed condition. This was confirmed by Eggers (1923) and Snodgrass (1924).

Mclndoo (1922) working on Johnston’s organ in the honey-bee showed that in these insects it took a peculiar form. He describes the scolopidia (about eighty in number in the worker) as attached at their distal ends to a series of chitinous knobs developed on the outer surface of the intersegmental membrane. These knobs are arranged in a ring around the outside of the base of the second segment of the funiculus (the pedicellus being the first). The intersegmental membrane is much strengthened by strands of flexible chitin, between which the scolopidia pass.

An investigation was made of the form that Johnston’s organ takes in the ant Formica rufa, the antennae of which are more amenable to sectioning than those of smaller species. The structure closely resembles that of the bee. The scolopidia, about twenty in number, are arranged in a ring around the pedicellus. Their proximal ends are attached to the wall of this segment. The distal end of each scolopidium is attached to a chitinous knob on the outer surface of the intersegmental membrane. These chitinous knobs are posterior projections from a narrower chitinous ring, which passes completely around the outside of the base of the second segment of the funiculus. The inner margin of the ring is set, tightly, in a corresponding groove in the wall of the second segment. The outer margin lies in the intersegmental membrane.

The intersegmental membrane is thick, and strengthened with bands of chitin. It is not very convoluted, however, when compared with that of other segments: this restricts the range of movement of the second segment upon the pedicellus, a fact which can be verified by observation of the relative mobility of the various segments.

The thick intersegmental membrane may have a twofold function. First, it prevents too wide a range of movement of the funiculus on the pedicellus, and secondly, it may take some of the strain across the joint between the two segments. Both of these activities would protect the scolopidia. It is generally accepted that the normal function of chordotonal organs, or scolopidia, is to respond to changes in tension. In Johnston’s organ changes in tension could be produced by slight changes in position of the funiculus relative to the pedicellus. Different orientations to gravity will result in different rotational forces on the funiculus. These will tend to displace the funiculus by different amounts; the actual size of the displacement is restricted by the intersegmental membrane, thus keeping it within the range of sensitivity of the scolopidia, and also preventing large displacements which might damage them.

The investigation of the ant’s force compass reaction to gravity showed that pairs of orientations at θ° and 360–θ° were equivalent to each other. This alone would not explain the occurrence of the four relationships obtained in the light/gravity substitution experiments. It was therefore decided to make a study of the ant’s reactions to light, with the object of ascertaining if equivalent orientations existed in the light compass reaction also. As the outward and return tracks of ants going to and from their feeding place are orientated at 180⁰ to each other, and since these tracks are both orientated relative to light from the sun (on occasion), it was thought that perhaps for the light compass reaction orientations 180⁰ apart might be equivalent to each other.

The experiments described here were very similar to those for the equivalent investigation of gravity orientation. The turntable described for those experiments was placed so that its surface was horizontal and could be rotated in a horizontal plane. The apparatus was set up in a dark-room, with an electric bulb (210 V. 60 W.) placed the far side of the room (about 3 m. away) on a level with the turntable. This bulb was suitably screened so that its beam of light was restricted to the turntable, and did not reflect from other objects in the room. The light could be switched on and off by a foot-operated switch. A dim, red safe-lamp was placed by the turntable so that the ants could be seen while the light was off.

A Perspex screen, on which to plot the tracks, was found unsatisfactory for these experiments, and the track of the ant was therefore plotted directly on the surface on which the ant walked. The surface of the table was covered by a sheet of cartridge paper, lightly pencilled into 1 in. squares. The track of an ant was plotted by marking lightly, in pencil, the points at which it entered and left a square: this being done when the ant was one square ahead. A fresh sheet of paper was used for each ant. In addition to the squares on the paper a diameter was lightly pencilled in every 30°, to assist in estimating the angle of orientation. The paper was attached to the turntable by four low mounds of Plasticine, which cast little shadow.

An ant was placed on the turntable and allowed to wander at random for 2 min. to become accustomed to the conditions. The escape reaction was then elicited by a gentle touch from a glass rod. The ant ran in a straight line, and its track was plotted. While it was running the angle of orientation 0° was estimated and the angle through which the turntable had to be turned was calculated. After the ant had travelled about 4 cm. the light was switched off, the table turned rapidly but smoothly through the calculated angle, and the light switched on again. The ant usually turned during and after the turning of the table, and then continued in a straight line. The track was again plotted. The track was then drawn in in detail, and the angle through which the table had been turned recorded. The whole procedure was then repeated. On an average five tracks were recorded for each ant.

The angles of orientation for each track (measured clockwise from the direction of the light) were then recorded before and after turning.

Three series of experiments were done:

• (a) Attempting to turn the table through 180—2θ°. This would turn an ant facing at 6° to head in a direction at 180—θ°.

• (b) Attempting to turn the table through 2θ°. This would turn an ant facing at θ° to head in a direction at 360—θ°.

• (c) Turning the table through 180°. This was done alternately clockwise and anticlockwise.

  N.B. θ° is the acute angle between track and light direction, and is always less than 90°. Both series (a) and (b) involved turning in both directions.

The details of the number of readings taken are given in Table 4.

Graphs were plotted, for each series of experiments, of angle track to light before turning/angle track to light after turning. No significant difference was observed between the results of series (b) and (c). These two series were therefore combined, and will in future be referred to as controls.

The graph for these control experiments is shown in the lower part of Fig. 3.

The control graph shows points grouped about the main regression line θ₁= θ₂, with some few other points scattered outside the main band. Some of these scattered points seem to lie about the line θ₁ = 360–θ₂, although the relationship, if it exists, is not a close one. If all the points obviously not in the main band are considered their scatter seems to be random. When the results corresponding to these points are examined it is found that in all but three cases the table was turned through a large angle, more than 160°. Turning the ants through large angles seems to disturb them rather more than small angles, and there is then a tendency for them to run rapidly straight on, rather than taking up a particular orientation. It is thought that this tendency is the reason why some of the scattered points lie along the line θ₁=360–θ₂, and that the arrangement is not due to any equivalence of orientations.

The conclusion drawn from the control experiments is that if the turntable is turned so that the ant is swung away from its original heading, it usually returns to its original orientation. As at least thirteen points out of sixty-five fall well outside the main band of points on the graph, the accuracy of the relationship was calculated by finding the area on either side of the fine within which 80 % of the points fell. No statistical calculation was made. The error in turning, made by the ant, calculated on this basis was + 15⁰. This indicates the possible inaccuracy of a fight compass reaction in this species (M. laevinodis).

The upper part of Fig. 3 shows the relationships for series (a) where the turntable was turned through 180—2θ°. The main regression of = is still found here, but in addition some points lie about the two lines represented by = 180—2θ °This last relationship would be the result of ants not turning after the table had been accurately rotated. In fact it was found that when the table had not been turned through the required angle the ants themselves often turned to take up a new orientation. This is shown in Table 5. Points are taken to show the relationship when they fall within 15⁰ of the regression line; when they fall near both lines they are allocated to the nearest.

These results show that if an ant is orientating at θ° to the light, and is turned toward or past a direction in which it would be heading at 180—θ°, it then turns either to renew its original orientation, or to take up a new orientation at 180—θ°. This indicates that the two orientations are equivalent in the light compass reaction, under the conditions used.

This equivalence is extremely surprising, for nothing in the normal behaviour of the ant would indicate such a result. Moreover, in the two equivalent orientations the right shines into the same compound eye, and nothing yet known about the structure of the eye or the central nervous system provides a basis to explain the phenomenon. It was therefore decided to investigate the structure of the ant’s compound eye, in case any structural linkage of ommatidia occurred, and also to check upon the accuracy of normal light-compass orientations.

Studies by Santschi (1923) and Werringloer (1932) have shown that the eyes of ants have a typical eucone structure. The rhabdome is very fine, and the number of retinal cells around each rhabdome is reduced to seven. Sections cut through the eye of M. ruginodis confirm this description.

The individual eye contains between two and three hundred ommatidia. The field of view of each eye is from directly forwards to within 10° of the body’s longitudinal axis behind. This means that there is a blind arc of 20° behind the ant. As an ant can run quite satisfactorily directly away from a lamp, it can presumably orientate by keeping light out of any ommatidium. The field of view in the vertical plane is from vertically upwards to about 43⁰ below the horizontal.

The angle between adjacent ommatidia in the same plane varies from 6° to 12⁰ with a mean value (from sixty-eight measurements) of 9⁰. The accuracy of the orientation to light was previously shown to be ±15°. In the experiments described the angle subtended by the light-source was 2°.

In experiments on the light-compass reaction with other insects (von Buddenbrock, 1931, 1935; von Buddenbrock & Schulz, 1933) a point source of light was always used. It was shown that corrective turning movements occur when the light moves from the originally illuminated ommatidium into an adjacent one. The conditions, however, do not allow us to distinguish between the two possibilities:

(a) That turning movements occur when light first shines into adjacent ommatidia;

or (b) That turning movements occur only when light has wholly left the original ommatidium.

In the experiments described here, the first possibility would lead to an accuracy of orientation of ± the largest ommatidial angle minus the angle subtended by the fight source, this figure being io°. The second possibility would lead to an accuracy of orientation of + the largest ommatidial angle plus the angle subtended by the light-source, this figure being 14°. This suggests that the second possibility is the more correct. More experiments on this subject are, however, needed.

The structure of the eye does not reveal any basis for the equivalence of two orientations. Neither does this equivalence have any manifestation in the ant’s normal behaviour; it is apparently brought forth only under the experimental conditions used. No explanation can as yet be given for it.

The sensitivity of Myrmica ruginodis to polarized light has previously been demonstrated (Vowles, 1950). It is now known that Lasius niger (L.) also possesses this sensitivity (Carthy, 1951). A histological examination of the eye of Myrmica ruginodis was made in order, if possible, to identify the insect’s analyser. The study revealed that no analyser was present either in the crystalline cone, or in the corneal lens. The analysing structure must therefore be in the rhabdome, the dimensions of which were too small to allow any satisfactory observations to be made.

The work of Menzer & Stockhammer (1951) shows that each rhabdomere acts as an analyser, and that in the eye of the bee the planes of analysis of the rhabdomeres are arranged tangentially about the main axis of the rhabdome in each ommatidium. This confirms the hypothesis of Autrum & Stumpf (1950), which was based on electrophysiological studies. There is no reason to suppose that the mechanisms of the ant’s eye differ from the bee’s. There are, however, only seven rhabdomeres to each ommatidium in the ant, and presumably the analysers are arranged in a heptagon around the axis of the rhabdome, rather than in an octagon as in the bee.

The experiments with M. ruginodis used only vertical beams of plane-polarized light. Such beams could stimulate only those ommatidia looking vertically upwards. In the ant, where the ommatidial angle is so large, only one ommatidium from each eye fulfils this condition. It appears therefore that ants can orientate to the plane of polarization of light when only one ommatidium in each eye is stimulated.

The angle between the sides of a heptagon is approximately 51⁰. Assuming the sides of the figure to be numbered consecutively 1-7, then the angle between side 1 and side 4 or side 5 is 27°. The range over which one retinula cell is more strongly stimulated than any other (when the plane of polarization most nearly coincides with its own plane of analysis) is therefore 27⁰. In the experiments with polarized light the accuracy of orientation was found to be ± 27⁰. While the exact coincidence of the two figures is unexpected, their closeness would lead one to suggest that corrective turning movements occur only after the plane of polarization has been rotated until the originally maximally stimulated retinula cell is no longer the most highly stimulated. This could be due, as von Frisch suggests, to a change in the pattern of stimulation of the ommatidium as a whole; there is, however, no evidence to support this hypothesis. It is equally possible that orientation to polarized light is performed by keeping the stimulation of a single retinula cell maximal; all that such a process requires is that the maximally stimulated cell is maintained so, by inhibiting any turning movements caused by stimulation of other retinal cells, until they are more strongly excited. This hypothesis would need a simpler neurological mechanism than does that of von Frisch, and the mechanism parallels that required for the light compass reaction.

If the above hypothesis is accepted it means that the ommatidium is not a functional unit, except in physico-optical terms. It should be stressed that the experiments so far performed on the fight compass reaction do not allow us to decide whether the retinula cells in a single ommatidium are acting in unison, or individually or successively. The possibility remains that even in the light compass reaction it is a single retinula cell which acts as the ‘fixation’ point in the eye. If this were so, then orientation to light and to polarized light would involve very similar mechanisms.

The main object of the experiments described in this paper was to investigate orientation to light and to gravity, in the hope that the results obtained would help to clarify the causes of the complexity of the correlation between successive orientations to these two stimuli. The experiments on geo-orientation show that an ant on a vertical surface may confuse orientations related by the equation θ₁ = 360–θ₂. This confusion, or equivalence, between two orientations does occur under natural conditions. It was suggested that an ant uses the georeceptor in only one antenna at one time, and that the taxis mechanism cannot distinguish between the information received from the two antennae.

In the light compass reaction of ants, under the experimental conditions used, there is a confusion, or equivalence, between orientations related by the equation θ₁= 180–θ₂. This confusion is not shown under natural conditions, and no explanation can be given for. it. It is suggested that, as in orientation to gravity, the taxis mechanism cannot distinguish between information received from the visual centres during two equivalent orientations.

It was suggested in a previous paper that compass reactions to light, gravity and polarized light were controlled by a common taxis mechanism, the ‘setting’ of which determines the actual direction taken relative to any of these stimuli. The mechanism, presumably, functions in response to information received from the sensory centres. Since the pair of equivalent orientations for gravity differ from the pair for light, the equivalencies cannot be due to defects in the taxis mechanism alone; for if they were, one would expect the same equivalencies for both stimuli. The defect must be in the information received by the taxis mechanism, and it is suggested that the two equivalencies for each stimulus occur because the information for two equivalent orientations is the same. Further, if the taxis mechanism cannot distinguish between information received from different sensory centres, then successive orientation to two different types of stimuli will be correlated in a complex way. Consider four examples of how this might happen for light and gravity:

1. An ant on a vertical surface is orientating at θ° to gravity, using the Johnston’s organ in its right antenna. The taxis mechanism is ‘set’ in the corresponding way.

   Light is now interchanged with gravity as the orientatory stimulus. The taxis mechanism maintains its ‘setting’, and therefore the ant turns until it orientates to the light at θ°. The successive orientations are therefore related by the equation θ₁ = θ₂.

2. An ant on a vertical surface is orientating at 360–θ° to gravity, but is using its left antenna: this is the equivalent orientation to (1) above. The taxis mechanism is therefore ‘set’ in exactly the same way as in (1) above.

   Light is again interchanged with gravity, and as the ‘setting’ of the taxis mechanism is the same as before, the ant again turns until it orientates at 6° to the light. The successive orientations are therefore related by the equation 360–θ₁ = θ₂.

3. An ant on a vertical surface is orientating at θ° to gravity, using its right antenna as in (1). The taxis mechanism is ‘set’ in precisely the same way as in the two examples already given.

   Light is now interchanged with gravity, and the taxis mechanism maintains its ‘setting’. However, a confusion now arises, so that it cannot distinguish between θ° and 180–θ°; so that the ant turns until orientating to light at the latter angle. The successive orientations are therefore related by the equation θ₁= 180–θ₂.

4. An ant on a vertical surface is orientating at 360–θ° to gravity using its left antenna, as in (2). The taxis mechanism is ‘set ‘precisely as in all the other examples.

Light is now interchanged with gravity as the orientatory stimulus. However, the confusion again arises, as in (3), between equivalent orientations, and the ant again turns to orientate at 180–θ°. The successive orientations are therefore related by the equation 360–θ₁= 180–θ₂, which becomes θ₁= 180 + θ₂.

The same relationships will be found when the ant orientates to light first. It will be seen therefore, that if the taxis mechanism does not distinguish between information from different sensory centres, or from the same centre when the ant orientates in equivalent directions, then successive orientations will be correlated in precisely the four ways demanded by the experiments on the interchange of light and gravity.

While the complex correlations following the interchange of light and gravity can be analysed as shown above, the results of interchanging light and polarized light cannot be so analysed. The experiments described here in no way clarify the complexity of the latter results. Consideration of the experiments on the interchange of light and polarized light suggests, however, the direction which further investigations might take: in the experiments there was in fact a double change of stimulus; for the light not only changed from normal to polarized, but also changed its direction from horizontal to vertical. It is not known what effect a vertical displacement of the light source has on an ordinary light compass reaction, although under natural conditions an ant can make good a general direction across a rough terrain, on which it continually rolls, pitches and yaws. Nor is the effect on orientation known of rotating the plane of polarization of a beam of light which is other than vertical, although bees can learn to orientate to such a stimulus. Until such effects have been analysed it would be unprofitable to speculate about the results from interchanging light and polarized light.

In all the experiments described here, the fine of action of the orientatory stimulus lay in a plane at righf angles to the dorso-ventral axis of the ant; variations in direction of the stimuli were always in this plane, as were the movements of the ant; thus the experiments were concerned only with the control of yawing. In all the observations made an ant never turned through more than 180°, either after change in direction of a single stimulus, or after interchange of different stimuli. This means that an ant always turned to its second orientation through the smallest angle possible: a similar result was obtained by von Buddenbrock & Schulz (1933). This implies that the sensory field has been divided into two functional halves, stimulation in one half causing turning in one direction, and stimulation in the other half causing turning in the opposite direction; the actual line of division corresponding to the line of action of the original stimulus. The maintenance of this functional division is presumably a property of the taxis mechanism.

Orientation to polarized light, however, differs from orientation to light and gravity in one important respect: for while the last two stimuli are vectorial in character, the plane of polarization is not. Thus, if the plane of polarization is rotated through 180⁰ the stimulus situation remains unchanged, although for light and gravity this is not so. The plane of polarization apparently lies right across the visual field, and cannot be said to act in any particular half of it. Therefore although all three types of orientation depend on keeping the sensory field constant, the method of analysing the field for polarized light must differ from that used in the light compass reaction. More work on this subject is still needed, but there is already good reason to suppose that the same taxis mechanism underlies all three types of orientation.

The work described was done in the Department of Zoology and Comparative Anatomy at Oxford, while the author held a Junior Research Grant from the Department of Scientific and Industrial Research.