Desert ants (Cataglyphis fortis) are scavengers, inhabiting the barren salt-pans of the North African Maghreb. In search for prey, these ants traverse their habitat on tortuous paths. In the absence of landmarks that could guide them [1], they head on a direct way towards their nest as soon as they have encountered a suitable booty [2,3]. The orientation mechanism that enables Cataglyphis to accurately return to their – mostly inconspicuous – nest entrance is path integration [4], i.e. the ability to combine the individual segments of an outbound journey in order to create a global home vector, which points back to the nest with the correct distance and compass direction.
One prerequisite for path integration is a reference system for the compass direction of a walked course. Cataglyphis utilizes both the position of the sun [5,6] and, in particular, the pattern of polarized skylight [7-10]. The other requirement of a path integrator is some form of odometer, which is probably implemented in desert ants as a kind of step counter [11,12], whereas self-induced optic flow and energy expenditure have been ruled out as the predominant sources of information [13-16].
Remarkably, the ants' path integrator still functions accurately when parts of the journey lead over a series of hills, thereby increasing the actual walking distance compared to the bee-line distance that is relevant for determining the home vector [16,17]. A three-dimensional outward journey, followed by a homebound trip carried out on level ground, causes no directional error [18]. This leads to the conclusion that the ants' path integrator correctly incorporates distances that were walked on slopes with their corresponding ground distance, not the actual walking distance. This ability to re-calculate a path length to its ground distance enables desert ants to orientate accurately in undulating terrain even if the path towards a food source and the way back to the nest lead through areas of different topography.
Based on these observations we now ask how sophisticated a desert ant's representation of its three-dimensional environment really is. As a starting point, we formulated three hypotheses that assume mechanisms of increasing complexity. These hypotheses lead to different predictions of how the ants would respond in test situations after specific forms of training (see Results):
Hypothesis A: The ants' path integration module works essentially in the horizontal plane; the ants store no information about the 3-D component of their environments. For this hypothesis we assume that distances walked on slopes are corrected to ground distance in real-time, but any additional information about the three-dimensionality of an itinerary is discarded. The world as perceived by the ant is a plane projection without vertical expansion.
Hypothesis B: The ants combine their 2-D path integration as outlined above with additional information about the three-dimensional structure of a walked path. E.g., motor commands for a vertical change of direction could be coupled to the current status of the home vector, or to specific landmarks. The ants could be prompted to execute these motor commands either in the sequence that was learnt on preceding foraging forays, or at specific distances from the end points of their trip. Thus, information about vertical changes of direction would be remembered in a form of procedural knowledge, which could already be demonstrated for 2-D itineraries [19]. If such changes fail to appear along a known route, this could affect the ants' homing behavior. This hypothesis is similar to the described orientation by local vectors that Cataglyphis uses in the horizontal plane. Here, the animals link local vectors [20] and motor commands [20,21] to familiar landmarks, which ultimately connect in sequence to complete routes defined by local vectors (shown in Melophorus bagoti: [22]).
Hypothesis C: The ants possess a path integration module that is able to compute a correct vector with respect to all three dimensions, i.e. a true 3-D vector. In principle, a 2-D representation of itineraries that corrects walking distances on slopes to their respective ground distance would be sufficient to navigate accurately between the nest and a food source. However, the desert ants' path integrator is prone to systematic errors [4], and locates its goal with an uncertainty that increases with the distance walked [23,24]. Under these circumstances, information could be valuable whether a food source is located at the top or the bottom of a slope. Sandy deserts are also an example of undulating terrain that is lacking unambiguous visual landmarks, which otherwise could help to home in on a known food source [25]. If the vertical component of such a 3-D vector cannot be followed, e.g. due to the lack of descent/ascent opportunities, this should affect the ants' homing behavior as well.
Unfortunately, it is difficult to discriminate experimentally between these hypotheses. In order to unequivocally prove, for example, that ants do compute a true 3-D vector, it would be necessary to perform the analogous experiments as in 2-D [4,26]. However, it is not feasible to offer an ant the same freedom to approach and to choose any points in 3-D space as it is possible in 2-D. For this reason, we had to reduce the complexity of the 3-D problem to experimentally manageable pieces. Although none of these tests by themselves allows for an unambiguous decision in favor of any of the outlined hypotheses, their combination allows for rather firm conclusions about the level of sophistication of Cataglyphis' orientation in 3-D space.
Answers to all your Biology Questions
