Polarized light use in the nocturnal bull ant, Myrmecia midas

Solitary foraging ants have a navigational toolkit, which includes the use of both terrestrial and celestial visual cues, allowing individuals to successfully pilot between food sources and their nest. One such celestial cue is the polarization pattern in the overhead sky. Here, we explore the use of polarized light during outbound and inbound journeys and with different home vectors in the nocturnal bull ant, Myrmecia midas. We tested foragers on both portions of the foraging trip by rotating the overhead polarization pattern by ±45°. Both outbound and inbound foragers responded to the polarized light change, but the extent to which they responded to the rotation varied. Outbound ants, both close to and further from the nest, compensated for the change in the overhead e-vector by about half of the manipulation, suggesting that outbound ants choose a compromise heading between the celestial and terrestrial compass cues. However, ants returning home compensated for the change in the e-vector by about half of the manipulation when the remaining home vector was short (1−2 m) and by more than half of the manipulation when the remaining vector was long (more than 4 m). We report these findings and discuss why weighting on polarization cues change in different contexts.


Background
Arthropods are known to derive compass information using the pattern of polarized skylight [1][2][3][4][5][6][7][8][9]. Polarized light comprises light waves in which the wave occurs along a single plane. Light scatters after entering the earth's atmosphere and becomes partially linearly polarized. This creates an e-vector pattern in the sky arranged in concentric circles around the sun or moon's position [10,11]. The e-vector in the overhead sky remains in a stable orientation pattern perpendicular to the direction of the sun/moon. This stability makes the sky's polarization During the forager's outbound trip to the foraging tree, a polarization filter was placed over the forager with the polarization e-vector rotated ±45°of the ambient e-vector. This filter apparatus was used in a previous study [6]. (b) Diagram of measurements collected during polarization filter test. Measurements were made using a compass application on a smartphone. Initial orientation routes were measured from the nest entrance (a) to when the polarization filter was centred over the forager (b). Initial route directions (a°) were calculated with the tree direction from the nest as 0°. The magnitude of angle a has been artificially enlarged in this diagram for clarity, with angle a averaging 4.42°across all conditions during testing. Exit orientations were measured from the centre of the polarization filter (b) to the exit location of the ant on the filter's edge (c). Route directions under the filter (b°) were calculated from the forager's initial route direction. Reorientations were measured from the forager's exit location from the polarization filter (c) to the forager's path 50 cm after exiting the filter (d). Reorientation route directions (c°) were calculated from the under-filter route direction.
phial. Each forager was offered a small amount of honey and was then stored overnight in the dark (9 h). Each collected ant was marked with a small amount of enamel paint to exclude previously tested individuals. Foragers were released at the base of their foraging tree in the pre-dawn twilight, which corresponds to the time at which they typically return home [25]. We followed each ant as it travelled to the nest, and placed the centre of the polarizer on the ant when it reached a distance of 4-6 m or 1-2 m from the nest. Therefore, both inbound and outbound foragers were tested at the same distance from the nest. Similar to the outbound tests, the e-vector axis of the filter was oriented either ±45°relative to the dominant ambient polarization pattern. We recorded the initial orientation, the exit orientation and the reorientation of each forager in the same manner as in the outbound tests (figure 1b). Foragers were then followed for the remainder of their inbound path to ensure they returned to the nest site.

Conflict between home-vector length and nest location
Here, we tested individual foragers close to their nest but with a large remaining vector. We achieved this by first following foragers from Nest 1 to the foraging tree (14 m) in the evening twilight and collected them in a phial as they reached the foraging tree. Just as in previous inbound conditions, these foragers were fed, marked with paint, held overnight and released in the pre-dawn twilight. We released the foragers on the route at the halfway point between the nest and the foraging tree (7 m). Released foragers were allowed to return to 1-2 m from the nest entrance where the centre of the polarizer was placed over the ant. As with all previous conditions, the e-vector axis of the filter was either ±45°relative to the ambient e-vector. Identical to previous conditions, we recorded initial orientation, exit orientation and reorientation for each forager (figure 1b). Foragers were then followed for the remainder of their inbound path to record their final destination.

Statistical analysis
Data were analysed with circular statistics [26,27]  designating the initial heading as 0°for each animal. The shift magnitude of each path was calculated by taking the mirror of the difference between the forager's initial path direction and the forager's exit orientation in each −45°condition. This calculation allowed us to compare path shifts in both directions in degrees. Foragers' shift magnitudes were compared between the ±45°and between the two distance groups using Watson-Williams F-tests. If shift magnitudes between the two groups do not differ, then it means both groups rely on polarized light to the same degree. A Pearson's correlation coefficient was used to test the association between the lunar phase (in per cent) and shift magnitude under the filter. Lunar phase data were obtained from calculations in the Astronomical Almanac (http://asa.usno.navy. mil).

Outbound ants at different distances from the nest
When the polarization filter was placed on an outbound ant at both testing distances, they initially stopped moving and then slowly began to move in a chosen direction. Most foragers would again stop as they reached the edge of the filter and performed visual scans before continuing on their chosen path. Ants did not pause after exiting the filter and continued on route towards their foraging tree.

Outbound foragers at the 1-2 m distance
When the polarizer was rotated ±45°, individuals paused after the polarizer was placed overhead. After this short pause, most individuals continued to the foraging tree (+45°, n = 18; -45°, n = 22; figure 2c closed circles); a minority of individuals in both conditions, however, turned back and retreated (defined as individuals that returned to within 30 cm of the nest entrance after exiting the filter) to the nest after the polarizer was placed over them (+45°, n = 8; −45°, n = 5; figure 2c open circles). Focusing on only those individuals that continued to the foraging tree, when the polarizer was rotated 45°to the left (-45°), the foragers' exit-orientations leaving the filter were to the left of their initial path direction (mean ± s.e.m.; Nest 1: θ = −18.26 ± 6.56°; table 1 and figure 3). This path change under the filter was significant (Watson-Williams F-test, Nest 1: F = 4.31, p = 0.04). When the polarizer was rotated 45°to the right (+45°), forager exit orientations were to the right of their initial path (mean ± s.e.m.; Nest 1: θ = 32.81 ± 6.4°; table 1 and figure 3) and this shift was also significant (Watson-Williams F-test, Nest 1: F = 12.29, p 0.01). After exiting the -45°rotated filter the foragers reoriented significantly to the right (Watson-Williams Ftest, Nest 1: F = 9.95, p 0.01, mean ± s.e.m. θ = 26.23 ± 6.73°; table 1 and figure 2c), and after exiting the +45°rotated filter the foragers reoriented significantly to the left (Watson-Williams F-test, Nest 1: F = 10.79, p 0.01, mean ± s.e.m. θ = -29.43 ± 5.87°; table 1 and figure 2c). While the number of individuals who retreated was insufficient for statistical analysis, foragers' exit orientations shifted as would be expected, either to the left (mean ± s.e.m. θ = -35.34 ± 26.60°) or to the right (mean ± s.e.m. θ = 37.14 ± 9.33°) of the nest entrance direction corresponding with manipulations in the polar filter (-45°or +45°, respectively). Shift magnitude was not significantly different between the -45°and +45°   Closed circles indicate individuals that continued on to the forging tree after testing. Open circles represent foragers that retreated once the filter was placed overhead and these individuals returned to within 30 cm of the nest entrance after testing. n, number of individuals; Ø, mean vector; r, length of the mean vector.
conditions (Watson-Williams F-test; 0.65, p = 0.43). When the -45°and +45°shifts were combined, the shift magnitude was also not significantly different between the outward heading ants of the two outbound testing conditions (Watson-Williams F-test, F = 0.17, p = 0.68).

Inbound ants at different distances from the nest
When foragers were released back at the base of their foraging tree during the morning twilight, they also initially paused for a brief period and scanned the environment without translation, before travelling in the nest direction. Nest-bound foragers typically paused again once the polarization filter was placed        above them, yet some individuals continued their forward movement. The same behavioural difference occurred at the filter edge, as some inbound foragers did not stop at the edge of the polarizer. After exiting the filter, all foragers continued on to the nest entrance and entered the nest.

Inbound foragers at the 4-6 m distance
When the polarizer was rotated to the left (-45°), the foragers' exit orientations were to the left of their initial direction of orientation (mean ± s.e.m.; Nest 1: θ = -41. 16

Conflict between home-vector length and nest location
Inbound ants with 14 m home vectors were displaced on the route but half way home and had to travel only 7 m to find the nest. The ability of these ants to detect a change in the pattern of the polarized light was assessed at 1-2 m from the nest entrance. When the polarizer was rotated to the left (-45°), the foragers' exit orientations were to the left of their initial direction of orientation (mean ± s.e.m.; Nest 1:

Lunar phase
Across all conditions, lunar phase was not associated with changes in shift magnitude of foragers while under the filter (Pearson's correlation coefficient, r = −0.127, p = 0.074).

Discussion
In this study, both inbound and outbound foragers changed their heading direction in response to changes in the overhead polarization pattern. In outbound foragers, we found that distance away from the nest did not influence the weighting foragers gave to this cue. Conversely, ants rely most on the pattern of the polarized skylight when they are returning home (inbound) and have a long accumulated vector (4-6 m).
In M. pyriformis, use of the polarization cue was tested only in outbound ants close to the nest [6]. Here, when the polarized filter was rotated by ±45°, ants changed their heading direction in the appropriate direction but by less than half of the rotation (-21.8°for -45°rotation; +14.1°for +45°r otation). In our study, the outbound M. midas ants at both 1-2 m and 4-6 m away from the nest (figure 2)  Both species appear to choose a compromise heading direction between the celestial and terrestrial compass cues during their outbound journey, and this appears to hold true at different distances from the nest.
In the inbound condition, we found that foragers of M. midas tested 1-2 m from the nest compensated for the change in the overhead e-vector by about half of the manipulation in their heading (-24.86°f or -45°and 19.73°for +45°, figure 3a). Interestingly, unlike in the outbound conditions, inbound M. midas ants tested at 4-6 m from the nest compensated for well over half of the e-vector manipulation in their altered heading (-41.16°for -45°and 34.13°for +45°, figure 3b). Such large compensation was also found in inbound foragers that had a 14 m home vector but were released half way home and tested close (1-2 m) to the nest (-35.77°for -45°and 39.42°for +45°, figure 4). This shows that inbound ants respond more to a change in the pattern of polarized skylight than outbound ants. These results imply that inbound ants weight polarization cues differently: ants with a longer home vector respond more to a change in the polarization pattern.
Our results suggest that foragers use both terrestrial and celestial cues, but the weighting of these cues appears to change with the ant's foraging context. In this study, M. midas foragers en route appear to weight vector cues in combination with the surrounding terrestrial cues, shifting their paths significantly under an altered polarization pattern. Yet when M. midas foragers are displaced to a local area with a vector direction conflicting with the surrounding terrestrial cues, individuals ignore the accumulated vector and orient using only the terrestrial cues [25]. Thus, it appears that nocturnal Myrmecia ants use the pattern of the polarized skylight only when the readouts of the celestial and terrestrial cues align. Furthermore, when there is a conflict between the two sources of compass cues, the celestial cues are suppressed or ignored [25]. This further implicates navigational context as a factor in celestial cue weighting. These behavioural differences may arise as foragers in the polarization experiment encounter no mismatch in cue sets before the polarized light filter, causing them to respond to the altered polarized light pattern while under the filter. Whereas after displacements off-route, foragers are presented with a mismatch between the familiar visual territory and their stored views, causing them to ignore celestial compass information when returning home [25]. The significant differences in shift magnitude in inbound foragers under different conditions were not predicted. Inbound foragers travelling from long distances (14 m, longest foraging route at this site) show larger shifts under the filter compared with individuals that forage in trees closer to the nest regardless of proximity to the nest. These disparities suggest greater weight is being placed on the polarization compass when in conflict with terrestrial cues in these foragers. It appears that the proximity of the nest tree at the test location, a potentially salient terrestrial cue, does not decrease the observed shifts in these far-foraging, long-vector individuals, implying that vector length clearly influences the weight given to the polarization pattern cue. These increases align with the hypothesis that with longer accumulated vectors, ants put more weight on these vector cues [28]. In our case, this difference in weighting persists even after a 9 h delay, with the direction of polarized light, linked to the position of the sun, having changed. These delay periods align with this species' foraging ecology as foragers typically spend this period on their foraging tree overnight [25]. Our results also align well with those from our previous M. midas study where only long-vector (more than 5 m) individuals show any evidence of orientation using path integration after distant displacement [25]. It may also be possible that, as tree fidelity has not been studied in this species, there may be some other difference between individuals that forage further from the nest site and those that forage at a nearer tree. These differences could include disparities in visual scenes encountered by these two foraging groups at Nest 1 or potentially even genetic differences between foragers travelling different distances to the nest. Further study into these behavioural choices is merited to untangle these possibilities.
Furthermore, it is interesting that the large shifts in long-vector inbound foragers are not seen in outbound foragers travelling to the same foraging tree. Vector memories in these outbound foragers are based on past foraging trips, whereas inbound foragers are using the vector memory of the current foraging trip [29,30]. This discrepancy may influence the weight these individuals give the vector cue. Unfortunately, as Nest 2 has since died, our field site currently has only one known nest with individuals foraging long distances (more than 5 m), making study of these differences difficult. Further study into this species and its use of celestial cues for navigation is warranted.
It is worth noting that the observed heading directions in both outbound and inbound foragers could be in part due to visual changes caused by the filter, independent of the e-vector rotation. Beyond the e-vector shift, light intensity levels are reduced, and there are changes in the visibility and salience of both celestial and terrestrial cues under the filter. These changes could alter the weighting of cues in this study compared to foragers navigating under natural conditions.

Conclusion
We show that both outbound and inbound M. midas foragers respond to changes of the e-vector orientation. Outbound ants compensate only partially to the change in polarized light, and this holds true at different distances from the nest. Inbound foragers with a longer home vector respond almost fully to the change in the pattern of the polarized skylight.
Data accessibility. Our data are collected in the electronic supplementary material. Authors' contributions. Experiments conceived and designed by C.A.F., A.N. and K.C. All data were collected by C.A.F.