Experimental setup
Experiments were performed between early July and early September in the years 2004 to 2006 on desert ants (Cataglyphis fortis) in their natural habitat, a saltpan area at 34.52°N, 10.53°E, near Maharès, Tunisia. The ants belonged to six different nests. Each animal was tested only once.
Ants were trained to visit a feeder filled with small pieces of watermelon and biscuit crumbs. Training and experiments took place in open aluminum channels (width & height of side walls: 7 cm; see [17]). A plastic enclosure surrounded the nest entrance and guided foraging ants into the affixed training channel. Fine grey sand was glued to the channel bottom in order to increase traction. The inner side walls were painted a matt grey to prevent possibly irritating reflections from metallic surfaces. The upper end of the walls was covered with smooth adhesive tape, impeding escape attempts. Except for experiment 3, the channels provided no visually contrasting elements that could be used as landmarks or would generate optic flow cues in walking ants.
Experiments 1 and 2
In 'reciprocal' experiments, ants were tested on their way home (experiment 1), or on their way out, towards the feeder (experiment 2).
Training paradigms
The ground distance from the nest to the feeder was 12 m in experiment 1, and 10.5 m in experiment 2. In both experiments, different groups of animals underwent three different training paradigms (Figure 1A–C). (i) Flat training took place in a straight horizontal channel. (ii) Ramp training utilized a channel that first led away from the nest horizontally, followed by an ascending ramp and an elevated horizontal channel. The distance of the ramp base from the nest entrance was 5.2 m in experiment 1, and 6 m in experiment 2. The length of the ramp was 150 cm in experiment 1, and 195 cm in experiment 2. The slope was 70 degrees in both cases. (iii) "Λ" training led the ants to a feeder at level ground. However, on their way they had to climb a ramp (located at the same ground distance as in ramp training), walk a short way horizontally (0.35 m), and descend a second ramp (same slope) back to ground level. The ramp lengths in (iii) corresponded to those in training (ii).
The differing dimensions in experiments 1 and 2 are owed to the possible combinations of used channel modules.
Test paradigms
In experiment 1, we tested the ants on their homebound run from the feeder back to the nest. After flat and Λ training, single ants were transferred to the test channel in a plastic vial filled with biscuit crumbs and released when they had taken up a morsel of food in their mandibles. In the case of ramp training, single ants carrying food were led into the adjacent test channel via a "switch" near the feeder. Neither procedure caused apparent irritations in the animals.
In the test channel, the ants encountered six "decision points" at which they could either continue to walk horizontally, or descend on a ramp (Figure 1D). All ramps used in tests were 150 cm long and had a slope of 70 degrees. The decision points were located at the following ground distances from the nest (in order of their encounter by a homebound ant): 10.5/8.9/7.3/5.7/4.2/2.6 m. "Decision points" were designed as a widening in the channel that led into two parallel channels of 7 cm width (Figure 1D, inset). One of the channels would continue horizontally, while the other led immediately to the descending ramp. On one experimental day, it would always be the channel on one side leading on horizontally at all 6 descent opportunities. The alignment was alternated between experimental days, but there were no significant differences in the results (p > 0.1 for all training paradigms, χ2 homogeneity test). In Homebound tests, animals were tested after sufficient training, i.e. when there was a steady flow of animals that approached the feeder unhesitatingly and at high speeds. Animals were marked after testing in order to exclude them from further experiments.
In experiment 1, we also tested ants in a flat channel (length: 14 m) laid out in parallel to the training channel ('flat control'). For this control experiment, we included a further ramp training paradigm. Here, the ramp was not located at 5.2 m, but at 9 m distance from the nest (see Results section).
In experiment 2, ants were tested on their outbound run from the nest to the feeder. A switch near the nest entrance was used to guide individual ants into the test channel, laid out in parallel to the training channel. In the test channel, six ramps offering a choice to ascend or to continue on level ground, were set up at 3/4.5/6/7.5/9, and 10.5 m distance from the nest (Figure 1E). Access to the ramps was made possible by small "gateways", whose sides (each 1.75 cm wide) led up the ramp, while an opening in the center (3.5 cm wide) allowed the ants to pass through underneath the ramp (Figure 1E, inset).
In this experiment, ants were marked individually with a three dot color code of acrylic paint on their thorax and gaster. An ant had to visit the feeder at least five times prior to being tested. This ensured that an ant leaving the nest was indeed heading for the feeder. This was not necessary in Homebound tests, as a food-carrying ant will always head for home, irrespective of the number of previous visits to a food source.
Experiment 3
Training paradigm
Ants were trained to walk through a horizontal channel to visit a feeder at a distance of 6 m from the nest entrance (Figure 7A). The channel was fitted out with a series of landmarks, similar to those used by Andel and Wehner [29]. These landmarks were 12 cm high, painted black on one side, and attached to the channel walls every 50 cm. By design, they were noticeable when passed on a homebound trip, while being rather inconspicuous on an outward journey. The landmarks should be associated by the animals with their homebound trip and provide an incentive to traverse the channel even when the route included a vertical diversion that was unknown to the animals, as was the case in the critical test (see below). Behind the nest and the feeder, we placed two "mock ramps" of 2 m and 1.5 m length, respectively. These ramps had no connection to the training channel, but provided a visual scenery, which matched that of the subsequent test situation. Prior to the training, ants were marked individually. Only ants that had visited the feeder at least ten times were tested, and each animal was tested only in one of the two test conditions, either the control or the critical test.
Test paradigms
Ants that were sufficiently trained were caught at the feeder and transferred to the test channel. Before their release, it was ensured that the ants had a biscuit crumb between their mandibles and thus were motivated to return to the nest.
In the control test, the channel consisted of a horizontal segment of 6 m length, laid out in parallel to the training channel (Figure 7A). It was fitted with the same landmarks as the training channel and ended at a test ramp (length: 2 m; slope: 70 deg). Behind the release point stood another mock ramp (1.5 m in length).
In the critical test, the release point was located in an elevated channel segment that led immediately to a descending ramp (length: 1.5 m; slope: 70 deg). At its base, the ramp connected with a horizontal channel. In 6 m ground distance from the release point, this channel ended at the base of a test ramp as in the control experiment (length: 2 m; slope: 70 deg). Both the descending ramp and the horizontal channel were again equipped with landmarks.
Data
Experiment 1 and 2
We recorded the path of ants at maximum for 2 minutes, or until they had made 10 U-turns in the horizontal channel or on the ramps. The recording of the ants' behavior also ended if they had covered the full length of one of the ramps, thus clearly indicating their choice for descent or ascent. In Homebound tests, we considered an ant's run also finished when it had reached the nest-ward end of the horizontal channel. Once ants had reached this dead end, they started to search in this part of the channel for a way out, but did not walk back to the last "decision point" or beyond.
We analyzed the frequencies of animals choosing to descend or ascend a test ramp for more than 20 cm (walking distance) in Homebound and Outbound tests, respectively. If, occasionally, an individual made more than one descent or ascent, only the first one was included in the analysis. These frequencies show whether different training paradigms had an influence on the general acceptance of sloped channel segments. Furthermore, we analyzed the distance that ants walked on a ramp before the first turn in their path. Thus, we obtained not only data about a general acceptance of ramps, but also information about the stretch of a slope that an ant intended to walk after the three different training situations.
In order to examine whether any of the six available ramps was preferred over the others, we calculated the frequency of ramp choice. This was done by summing up the full descents/ascents on a ramp, as well as the sum of passes that ants had carried out at the respective "decision point". In this instance, we considered as a valid run only descents/ascents that covered the full length of the ramps. If ants turned around prior to the end of the ramp, we took this as an indication that they had ultimately decided against climbing up or down a slope at this position.
We ensured that this strict criterion for analyzing descents and ascents did not mask any results by additionally examining our data with a "softer" criterion: In this, we looked at the initial ramp choice by considering the first ramp on which the animal descended or climbed for more than 20 cm, irrespective of later turns. These results are provided in the additional files section of this article.
In the flat controls of the Homebound test, we noted the first ten turns in the ants' walked path.
Experiment 3
In both test situations, we recorded the length of the first ascent that an ant undertook on the test ramp. In the critical test, some ants turned around several times on the descending ramp before reaching the level segment. These turns had no influence on the length of ascent on the test ramp (p > 0.5; N of turning ants = 16; N of not turning ants = 21; Mann-Whitney U-test).
Statistical analysis
Experiment 1 and 2
Frequencies of animals choosing to descend or ascend, as well as frequencies of chosen ramps, were analyzed using the χ2 homogeneity test. If descent/ascent frequencies showed significant differences, these were localized with a pair-wise comparison using Fisher's exact test and Bonferroni correction for multiple comparisons. In the analysis of ramp choice, deviations from homogeneity were located by using the squared standardized residuals. The lengths of descents and ascents following the three training paradigms were compared for each experiment using the Kruskal-Wallis H-test. Differences between pairs of sample groups were localized using the Games-Howell post-hoc test for pair-wise comparisons. In the flat controls of the Homebound test, we used the first U-turn as an indicator for the position of an ants' search area. We compared these data between the different training paradigms using the Kruskal-Wallis H-test and Games-Howell post-hoc test.
Experiment 3
The length of ascents on the test ramp in the critical test (with an initial descent) and the control test (which had no descent) was compared using the Mann-Whitney U-test.
All statistical analysis was carried out using SPSS for Windows, version 12.0.1.
The experiments comply with the "Principles of animal care", publication No. 86-23, revised, 1985 of the National Institute of Health, and with the current laws of Germany and Tunisia.
Answers to all your Biology Questions
