QUICK SEARCH:   [advanced] Author: Keyword(s):
Home     Help     Feedback     Subscriptions     Archive     Search     Table of Contents

This Article Summary Full Text Full Text (PDF) Alert me when this article is cited Alert me if a correction is posted Services Email this article to a friend Similar articles in this journal Similar articles in PubMed Alert me to new issues of the journal Download to citation manager Citing Articles Citing Articles via HighWire Citing Articles via Google Scholar Google Scholar Articles by Graham, P. Articles by Collett, T. S. Search for Related Content PubMed PubMed Citation Articles by Graham, P. Articles by Collett, T. S.

View-based navigation in insects: how wood ants (Formica rufa L.) look at and are guided by extended landmarks

Paul Graham^* and Thomas S. Collett

School of Biological Sciences, University of Sussex, Brighton, BN1 9QG, UK

View larger version (25K): [in a new window]   Fig. 1. Routes taken by experienced ants along the wall. Top: sketch of the arena and wall. (A,C,E) Mean path from start (S) to food (F), with 95% confidence intervals, for the 20 cm, 30 cm and oblique training routes respectively (20 cm route, n=154, N=17; 30 cm route, n=65, N=6; oblique route, n=88, N=6; incompletely recorded paths are excluded. n, number of trajectories; N, number of ants). The mean path was calculated by averaging, over all runs, the perpendicular distance from the wall (y) at 10 cm intervals along the wall (x). (B,D,F) Trajectories from three different ants performing the 20 cm, 30 cm and oblique routes respectively. Data are rotated to align the wall positions. (G) The arena was divided into six sectors, and runs from all conditions were separated into six groups depending on which sector contained the wall. To test whether ants behave uniformly within the arena, we assessed the quality of each run by determining how much each trajectory strayed from the direct route to the goal. This metric, referred to as normalised error, is defined as the mean difference in y value between the actual route and the direct route, normalised by the y value for the direct route. There are no significant differences between the six groups (one-way ANOVA, F=0.391, d.f.=5,286, P=0.86). (H) There is also no significant clustering of bad trajectories. Bad trajectories are defined as those that fall more than 2 S.D. from the mean, for straightness (Sinuosity; outer circle) or mean distance from the wall (Mean y; inner circle). Neither distribution differed significantly from random (Rayleigh test; mean distance from wall, r=0.18, P>0.3; straightness, r=0.18, P>0.3).

View larger version (33K): [in a new window]   Fig. 2. Responses of ants to the angular height of the wall. (A,C) Cumulative plots of mean headings, with 95% confidence intervals, for different wall heights and starting positions. Ants were trained on the 20 cm route with a wall that was 20 cm high and 120 cm long with the food 100 cm in the x direction. They were tested with different starting positions and wall heights. In A, ants were presented with the wall as in training and were released at a point either 20 or 40 cm from the wall. In C, the wall height was increased to 40 cm, and ants were released at 20 or 40 cm from the wall. Asterisks show where the mean path departs significantly (*P<0.05, **P<0.005) from a course parallel to the wall (dashed lines). (B,D) Individual paths plotted in terms of the angular height of the nearest part of the wall to the ant against the ant's distance along the wall. Traces stop at the point where the ant turned back towards the start position. The dotted line at 45° represents the angular height of the wall in the training condition. In B, the wall is 20 cm high, with a start position 40 cm from the wall. In D, the wall is 40 cm high, with a start position 20 cm from the wall.

View larger version (26K): [in a new window]   Fig. 3. Looking patterns while travelling along the wall. (A,C,E) Distributions of looking distances, as defined in the top diagram, for ants trained to the 20 cm, 30 cm and fixed routes respectively. Normalised frequency is the number of video frames (frame rate 50 Hz) corresponding to the condition of each bin divided by the total number of frames in the sample. For every 20 ms time step, we calculated the point at which the forward extension of the ant's longitudinal axis intersects the wall. The looking distance is defined as the difference between this intersection point and the current x position of the ant. Bins are 2 cm wide, and the centres of the modal bins are 21 cm for the 20 cm condition, 23 cm for the 30 cm condition and 11 cm for the fixed wall condition. (B,D,F) Distributions of the orientations of body axis for ants in the same conditions as above, again normalised by the total number of frames. At 0°, an ant's longitudinal axis is parallel to the wall and positive angles are clockwise (the ant faces towards the wall). The grey shaded area gives the distribution of orientations for those frames in which the wall is viewed with the frontal retina. The black shaded area gives the subset of this distribution for which looking distance is between 10 and 30 cm (dashed lines on A and C).

View larger version (21K): [in a new window]   Fig. 4. The fine structure of the ant's wiggly paths. (A) Each cycle of the ants' wiggles was split into quadrants according to the ants' orientation and turning direction. (i) Orientation away from the wall (-ve), clockwise turn (+ve). (ii) Orientation towards the wall, clockwise turn. (iii) Orientation towards the wall, counterclockwise turn. (iv) Orientation away from the wall, counterclockwise turn. (B,C) The mean distance travelled in each quadrant for the 20 and 30 cm conditions respectively. The large arrow represents the direction of travel. (D) Mean angular velocity in each quadrant plotted against orientation. The data are from 20 cm training routes.

View larger version (21K): [in a new window]   Fig. 5. A model of correction. (A) Model trajectories were generated with a sine function scaled to approximate the characteristics of real trajectories. The tangent to the curve gives the orientation of the model ant, and the apparent height of the wall is defined as the vertical angle subtended by the portion of the wall that is viewed by the frontal retina of the model ant. The clockwise scanning phase, during which the ant turns towards the wall, ends when the apparent height of the wall reaches the stored value, and the counterclockwise scanning phase ends at a set angle. With training conditions, a starting y distance of 20 cm from a 20 cm high wall, the trajectory is parallel to the wall. With a 40 cm high wall and a starting y distance of 20 cm, the view on the frontal retina matches the required retinal height at a reduced deviation from the straight-ahead direction. Since rotation in the counterclockwise phase is unchanged, the trajectory veers away from the wall. Conversely, with a 40 cm y distance from a 20 cm high wall, the model ant rotates further towards the wall to match retinal height, and its course veers towards the wall. (B) The scanning model was tested in experiments in which ants were guided along the middle of a 60 cm wide channel by two 20 cm high walls. (C-E) Plots of the orientation at the end of each scan cycle for the training wall and the low (15 cm) and high (40 cm) test walls. The shaded areas in D and E indicate the predicted mode for the distribution based on the scan model and on the mode of the distribution for training runs in C.

View larger version (20K): [in a new window]   Fig. 6. Responses to stepped walls. (A) The dimensions of the training and test walls from an experiment in which ants were trained to a 20 cm high wall and tested with a stepped wall. (B) Plots of mean headings during training (black lines) and under test conditions (grey lines). (C) Mean differences between test runs and training runs. Each test run is paired with the previous training run from the same ant. Error bars show the 95% confidence interval of the mean. The significance of these differences is calculated using a two-tailed t-test, tested against zero. Significant results are highlighted on the graph; **P<0.005, *P<0.05. Means are calculated from 22 test and 22 control paths from five ants. (D) In a second experiment, ants were trained to a stepped wall that began at a height of 40 cm and dropped to a height of 20 cm after 100 cm, and they were tested with a uniform 40 cm high wall. (E) Mean trajectories from test and training runs in this condition, and (F) mean difference between a test run and the previous training run by that individual. Means are calculated from 55 test and 55 control paths from seven ants. The arrow in F indicates the only point where the gradients of adjacent sections are significantly different (t-test, t=2.4, d.f.=108, P<0.05).

View larger version (22K): [in a new window]   Fig. 7. Stopping points. (A) Example trajectories from the control condition with no food present (food in training is 100 cm along the wall). (B-D) Stopping points from individual runs with walls of different length. The stopping point of a trajectory is taken as the point where the ant turns through more than 90° and heads back towards the start position (see Materials and methods). (B) Control condition with same wall length as in training, (C,D) Tests with an 85 cm long wall and a 155 cm long wall, respectively, putting visual cues and cues from distance walked into conflict. Individual stopping points are shown as crosses. Mean stopping points are shown by the intersection of the major and minor axes of the 95% confidence ellipse of the mean. Large arrows show the predicted stopping points along the x dimension (the predicted stopping points for visual cues are indicated by black arrows, the predicted stopping points for cues from distance walked by grey arrows). (E) Example trajectories from the 155 cm length wall condition. Trajectories remain at 20 cm from the wall until the start of search behaviour.