Cataglyphis fortis · two measurements inside one home vector
The Distance in Its Legs
An ant walks out on one body and comes home on another. Change its stride after the memory is written and the point where it starts searching moves. Then cover the view beneath its eyes and a second distance memory appears, still charged after the first has run out.
Direction is not the problem here. A celestial compass keeps the test channel aligned toward home. The question is smaller and stranger: how far? The first instrument encodes the outbound route as a number of effective strides. The second opens the raw turning records and asks what survives when ventral optic flow is withheld, restored, or never written.
Write the journey. Change the legs.
The ants learned a 10 m nest-to-feeder route on normal legs. In the first test, their legs were changed only at the feeder. Choose stilts or stumps below. Then move the timing to before both journey legs and watch the error cancel.
The stride register
model output in gold · 2006 group median in blue
This is not a fitted correction. It multiplies the selected stride wherever that stride occurs, so it cancels when both journey legs match. Keep 1.000 for the measured stride-ratio prediction. The paper's individual body-size and full speed normalization cannot be reconstructed from its summary, so its 14.25 m corrected stilt result is shown below only as an anchor.
Normal legs encode and discharge the same stride length. The model returns 10.00 m.
The direction is right while the distance runs long or short. In the 2006 first test, the published search-centre medians were stilt 15.30 m, normal 10.20 m, and stump 5.75 m, with 25 ants in each group. After the same modified ants later made both outbound and inbound trips on their new legs, medians returned to 10.55 m on stilts and 10.25 m on stumps.
That cancellation is the clean part. The exact quantitative fit is not. Measured mean stride ratios alone predict only 11.38 m on stilts and 6.62 m on stumps. Body size and speed change stride length too, and the stilt preparation slowed rather than accelerated the ants. The article reports a separately normalized and speed-corrected stilt prediction of 14.25 m. It does not publish enough individual-level inputs to rebuild that correction here.
| condition | when changed | raw-ratio model | observed median | observed IQR width |
|---|---|---|---|---|
| normal | neither | 10.00 m | 10.20 m | 2.40 m |
| stilts | after outbound | 11.38 m | 15.30 m | 3.24 m |
| stumps | after outbound | 6.62 m | 5.75 m | 1.81 m |
| stilts | before both | 10.00 m | 10.55 m | 1.45 m |
| stumps | before both | 10.00 m | 10.25 m | 1.76 m |
A second memory waits behind the paint.
In 2018, ventral eye regions were temporarily covered to remove the optic-flow contribution during selected phases. The stride integrator remained available. Every button below loads one experimental group from the authors' raw workbook.
Charge, discharge, restore
0 = ventral view open · 1 = ventral view blocked
Optic flow was charged outbound, blocked until the first turn, then restored. The later search moved to 15.80 m.
The latent charge
In 0-10, the first turn arrives at 7.65 m while optic flow is blocked. Restore the ventral view and the ants resume homebound travel. Their later per-ant search medians centre at:
15.80 m
The visual distance memory was unavailable at the first turn but was not erased. This is behavioural evidence for separately retained distance information, not a located neural reservoir.
Break the fixed-additive story
The 0-11 centre, 5.70 m, suggests a stride weight of 0.57 and an optic-flow weight of 0.43 if one forces a fixed additive reading. Keep that 0.43 weight after vision returns in 0-10 and the predicted second centre is:
11.95 m
Raw 0-10 centre minus that prediction: 3.85 m. A fixed weight misses. Context-dependent weights and competition both remain possible.
The strongest result is not that an ant counts. It is that one walking animal can carry distance in at least two cue systems, let one control the moment a search begins, and later reveal information retained by the other. The 2018 authors call their quantitative account rough. Their data do not decide whether the memories change weight with context, compete, or interact by some other rule.
The check: an instrument with teeth
The browser reruns these assertions from its embedded source table. The standalone verifier implements a second median route and also string-matches the visible fallback figures.
- ✓Independent stride routes agree: ratio arithmetic and explicit stride-count arithmetic both give 11.38 m for stilts.
- ✓Closed-form anchor: using the same stride outbound and inbound returns exactly 10.00 m.
- ✓Published raw-data anchors: all six 2018 group medians reproduce to the printed precision; maximum discrepancy 0.00 m.
- ✓Control on the control: shifting every 0-10 turning point by 4 m is rejected against the 15.80 m anchor.
Free choices: the effective-stride sensitivity multiplier; the alternative centre summary; and the quartile convention. Uncertainties: the 2006 individual body-size and speed correction cannot be rebuilt from published summary data; the 2018 1-11 row count conflicts with the figure caption; raw IQR endpoints depend on convention; and the interaction rule between odometers remains open. No Kruskal-Wallis statistic is recomputed: the 2006 raw searches are unavailable, and the 2018 paper reports selected pairwise tests rather than that omnibus test.
Run node research/the-distance-in-its-legs/verify-the-distance-in-its-legs.mjs.
What a search centre is, and is not
The 2006 centre was the median of the first six turns in one ant's nest-search path. The 2018 main centre used here is each ant's median of turns 2 through 6, followed by the median across ants. The first turn is also shown because it marks the switch from a relatively straight homebound run to search. Neither statistic is a claim about belief or conscious judgement.
Methods involving live animals
The completed 2006 study shortened legs or attached pig bristles to lengthen them. The completed 2018 study used temporary paint over ventral eye regions. This page describes those published interventions factually and provides no instructions for repeating them.
Sources and data pin
Wittlinger, Wehner, and Wolf (2006), The Ant Odometer: Stepping on Stilts and Stumps. Wolf, Wittlinger, and Pfeffer (2018), Two distance memories in desert ants: Modes of interaction. The supplementary workbook is pinned in the research README by SHA-256. The page makes no outside requests.