Compensation to visual impairments and behavioral plasticity in navigating ants

Sebastian Schwarz, Léo Clément, Lars Haalck, Benjamin Risse, Antoine Wystrach

This research paper, "Compensation to visual impairments and behavioral plasticity in navigating ants," explores the fascinating intersection of rigid neural architecture and flexible behavior in desert ants (Cataglyphis velox).

TL;DR

Desert ants are "fundamentally binocular" navigators; they cannot recognize a familiar route if one eye is covered, nor can they recognize a monocularly-learned route once both eyes are open. However, they exhibit remarkable plasticity by re-learning their entire environment from scratch in just a few hours, storing these new "one-eyed" memories in parallel with their original "two-eyed" ones.

What Problem Does This Paper Address?

In the world of robotics and biology, "robustness" usually means a system can continue to function despite damage. If a human loses sight in one eye, they can still recognize their living room. We assume biological systems possess this kind of "inter-ocular transfer."

However, insects are often viewed as more "hard-wired." The authors sought to understand how ants integrate information from two compound eyes and whether they can compensate for sudden visual impairment. Specifically, they investigated if ants possess the flexibility to recover navigational functions after losing half their visual field, and what that recovery reveals about how memories are stored in the insect brain.

Methods

The researchers conducted field experiments in Seville, Spain, using Iberian desert ants. The core methodology involved:

  • Eye-Capping: Using opaque enamel paint to cover one of the ant's two compound eyes.
  • Navigational Challenges: Ants were trained on an 8-meter route between a feeder and their nest.
  • Vector Manipulation: They tested both "Full-vector" ants (who have an internal compass/odometer) and "Zero-vector" ants (who must rely entirely on visual landmarks) to isolate the role of visual memory.
  • Learning Ontogeny: They observed "learning walks"—convoluted pirouettes ants perform to memorize a new area—to see if impaired ants would revert to "infant" learning behaviors.
  • Tracking: Paths were recorded using a combination of manual grid-mapping and a custom video tracking system developed by the Risse group.

Datasets

As a biological study, this paper does not use standard machine learning benchmarks. Instead, it generates a primary behavioral dataset consisting of:

  • High-resolution trajectories of Cataglyphis velox under various visual conditions (Sham control, Just-capped, and Compensated).
  • Comparative data between "In-and-Out" (full route experience), "Out-only," and "In-only" training groups.
  • The study utilizes a novel video tracker system (referenced as the "Risse group" system) to digitize paths in natural, cluttered environments.

Key Insights and Ideas

  • The Binocular Constraint: Ants do not possess inter-ocular transfer for route memories. A view learned with two eyes is a completely different "file" than a view seen with one. Adding or removing visual input acts as a "password change" that locks the ant out of its own memory.
  • Behavioral Plasticity vs. Neural Rigidity: While the memory retrieval is rigid (requiring a specific binocular match), the ant’s behavior is incredibly flexible. When the world looks "wrong" due to the eye-cap, the ant spontaneously restarts its "learning ontogeny"—the same process a naive ant uses when first leaving the nest.
  • Parallel Memory Storage: Re-learning the world with one eye does not overwrite the old two-eyed memories. When the eye-cap is removed, the ant can immediately navigate using its original binocular memories, proving the brain can store multiple "versions" of the same route.
  • The Coincidence Detector Model: The authors propose a neural model where Kenyon Cells (KCs) in the mushroom bodies act as coincidence detectors. They likely require simultaneous, specific input from both eyes to fire. If one eye is blocked, the "pattern" is broken, and the ant perceives the environment as entirely novel.

Key Results

  • Immediate Failure: 100% of "just-capped" ants failed to navigate home initially, veering off-course an average of 3.42 times per trip.
  • Rapid Recovery: Within just a few hours of "re-learning" (performing learning walks), eye-capped ants recovered near-normal navigational efficiency.
  • The "Uncapping" Paradox: Ants that learned a route only while monocular were unable to navigate that same route when their second eye was uncovered. They were "blinded" by the extra information.
  • Learning Walks are Essential: Ants prevented from performing learning walks (the "In-only" group) showed zero improvement in homing ability, even after hours spent outdoors.

Limitations and Future Work

  • Neural Verification: The proposed "coincidence detector" model is based on behavioral observation. Direct electrophysiological recordings of Kenyon Cells in monocular vs. binocular ants are needed to confirm the mechanism.
  • Sensory-Motor Decay: The ants showed a persistent motor bias (turning toward the open eye) that lasted longer than the memory recovery. The paper notes that this "sensory-motor defect" eventually fades, but the timeline and mechanism for this specific compensation remain "data in preparation."
  • Species Generality: It remains to be seen if this binocular dependency is universal across all insects or specific to high-visual-acuity foragers like Cataglyphis.

Reference: Schwarz, S., et al. (2023). Compensation to visual impairments and behavioral plasticity in navigating ants. bioRxiv preprint.

Summary updated

Abstract

Desert ants are known to rely heavily on vision while venturing for food and returning to the nest. During these foraging trips, ants memorize and recognize their visual surroundings, which enables them to recapitulate individually learned routes in a fast and effective manner. The compound eyes are crucial for such visual navigation; however, it remains unclear how information from both eyes are integrated and how ants cope with visual impairment. Here, we manipulated the ants' visual system by covering one of the two compound eyes and analyzed their ability to recognize familiar views. Monocular ants showed an immediate disruption of their ability to recapitulate their familiar route. However, they were able to compensate for this nonnatural impairment in a few hours by engaging in an extensive route-relearning ontogeny, composed of more learning walks than what naïve ants typically do. This relearning process with one eye forms novel memories, without erasing the previous memories acquired with two eyes. Additionally, ants having learned a route with one eye only are unable to recognize it with two eyes, even though more information is available. Together, this shows that visual memories are encoded and recalled in an egocentric and fundamentally binocular way, where the visual input as a whole must be matched to enable recognition. We show how this kind of visual processing fits with their neural circuitry.
BibTeX
@inproceedings{schwarz2024compensation,
  author = {Sebastian Schwarz and Léo Clément and Lars Haalck and Benjamin Risse and Antoine Wystrach},
  title = {{Compensation to visual impairments and behavioral plasticity in navigating ants}},
  booktitle = {Proceedings of the National Academy of Sciences},
  year = {2024},
  doi = {10.1073/pnas.2410908121},
}
Figure