Intelligent Design Icon Intelligent Design
Life Sciences Icon Life Sciences

Navigation Ability Crosses Phylum Lines — And That’s a Problem for Evolution

Evolution News
Photo credit: josullivan.59 @ Flickr

Scientists are funny. They can think up the most creative questions to test by experiment, like, “Can goldfish drive a car?” Laypeople joke that “A woman needs a man like a fish needs a bicycle,” but scientists rigged up an experiment to see if fish could drive a motorized wagon. Answer: apparently so! Fish have the gift of navigation, which leads to our point: the similar abilities of vastly different organisms to navigate remains a challenge to explain by evolution.

Scientists at Ben-Gurion University of the Negev announced the results of their test:

Are animals’ innate navigational abilities universal or are they restricted to their home environments? Taking the premise to the extreme, the researchers designed a set of wheels under a goldfish tank with a camera system to record and translate the fish’s movements into forward and back and side to side directions to the wheels. By doing so, they discovered that a goldfish’s navigational ability supersedes its watery environs. [Emphasis added.]

Yes, that is kind of adorable. It took a few days for the fish to learn to drive, but by the end of the experiment, they could aim at a target for a treat by bumping their little noses against the glass. Additional tests ruled out luck, and showed that the fishes’ navigational ability was real.

The study led the researchers to two conclusions. “The study hints that navigational ability is universal rather than specific to the environment. Second, it shows that goldfish have the cognitive ability to learn a complex task in an environment completely unlike the one they evolved in.

Maybe they didn’t evolve this ability. Maybe they were endowed with it by design. That would make sense after reading Eric Cassell’s book, Animal Algorithms. His conclusion of intelligent design as the source of navigation keeps getting support from new scientific papers.

Universal Search Strategies

If you were following tracks in a sandy prairie but lost the trail at some point, what could you do to find it again? Using your reasoning, you might backtrack to the last known track and then try what is called a “casting,” heading at an angle left then right while moving forward, hoping to cross the trail again. Many animals have additional sensory aids, like olfaction, to contribute to this strategy (think of a bloodhound or hunting dog following a scent). The odd thing is that completely unrelated animals share this ability. Fish and dogs and humans are all vertebrates, but what about ants? 

In PNAS, Reddy, Shraiman, and Vergassola write about their research on “Sector search strategies for odor trail tracking.” They begin by noting this near-universal ability across phylum divisions:

Ants, mice, and dogs often use surface-bound scent trails to establish navigation routes or to find food and mates, yet their tracking strategies remain poorly understood. Chemotaxis-based strategies cannot explain casting, a characteristic sequence of wide oscillations with increasing amplitude performed upon sustained loss of contact with the trail. We propose that tracking animals have an intrinsic, geometric notion of continuity, allowing them to exploit past contacts with the trail to form an estimate of where it is headed. This estimate and its uncertainty form an angular sector, and the emergent search patterns resemble a “sector search.”

How could such an ability evolve? In Darwinian terms, these animals came from different Cambrian phyla that never shared genes for body plans or brains. The authors do not present an evolutionary theory. Instead, they try to explain it as an emergent property of general intelligence and the geometric constraints of the system. By building computer models with certain constraints, they feel that casting behavior naturally emerges.

We use ideas from polymer physics to formulate a statistical description of trails and show that search geometry imposes basic limits on how quickly animals can track trails. By formulating trail tracking as a Bellman-type sequential optimization problem, we quantify the geometric elements of optimal sector search strategy, effectively explaining why and when casting is necessary. We propose a set of experiments to infer how tracking animals acquire, integrate, and respond to past information on the tracked trail. More generally, we define navigational strategies relevant for animals and biomimetic robots and formulate trail tracking as a behavioral paradigm for learning, memory, and planning.

This is all fine and good, but it glosses over requirements for this ability. Non-living objects, obviously, do not perform casting as a strategy. It might be imagined that a stream crossing dry ground might cast this way and that to find pathways forward, but that is not a strategy; it just happens by natural laws. A search strategy requires a being with sensors, memory, and the ability to process information. That’s an animal “algorithm” — watch them use the word:

In conclusion, we show that an optimized sector search strategy based on the memory of two or more recent detection events yields an oscillatory search path with increasing amplitude that naturally unifies the observed low-amplitude “zigzagging” and larger-amplitude “casting” behaviors into the same quantitative framework. This framework elucidates the geometric and computational constraints faced by tracking animals and identifies general features of the algorithms that efficiently solve the task, which can also be implemented for robotic applications. Insights and predictions developed here impact and should inform the design and analysis of future animal behavior experiments.

Animals — from mammals to sea turtles to ants with pinhead sized brains — possess the minimal requirements for the algorithm to work. If the trail has olfactory cues, follow it. If olfactory cues are lost, refer to a memory map of the trail. If there is no memory, slow down and start casting. If all else fails, go back to the last known point and start a new search. This shows that navigation requires more than just sensors and strategies, but the ability to prioritize and optimize them with calculations. This is good for designers to know.

In contrast with chemotaxis-based algorithms, we propose an alternative framework built on the searcher exploiting past contacts with the trail to maintain an estimate of the trail’s local heading and its uncertainty. A minimal memory of the approximate locations of the two most recent contacts suffices to delineate an angular sector of probable trail headings that radiates from the most recent detection point. The resulting “sector search” provides a quantitative description of trail-tracking behavior that unifies its various phasesand yields specific experimental predictions.

The authors show how to write the optimization in mathematical form. How does an animal do this? They don’t say.

Disambiguating Cues

At the level of neurons, animals need to distinguish between external motions of objects and internal motions of the body: e.g., the orientation of the head. Researchers at Sainsbury Wellcome Centre (SWC) for Neural Circuits and Behaviour at University College London (UCL) studied how brain cells sort this all out.

To navigate successfully in an environment, you need to continuously track the speed and direction of your head, even in the dark. Researchers at the Sainsbury Wellcome Centre at UCL have discovered how individual and networks of cells in an area of the brain called the retrosplenial cortex encode this angular head motion in mice to enable navigation both during the day and at night.

Mice and other mammals have vestibular signals from the inner ear that sense body motion. The cues from those organs are encoded by Angular Head Velocity (AHV) cells. By placing mice on rotating platforms as moving lines were flashed on a screen, the researchers were able to disambiguate the signals from internal body motions and those from external motions. When the lights are on, the team found, the mice’s navigation becomes “significantly more accurate,” implying that the brain knows how to switch to the more reliable cue when it is available.

“While it was already known that the retrosplenial cortex is involved in the encoding of spatial orientation and self-motion guided navigation, this study allowed us to look at integration at both a network and cellular level. We showed that a single cell can see both kinds of signals: vestibular and visual…. It’s pretty compelling that both the coding of head motion and the mouse’s estimates of their motion speed both significantly improve when visual cues are available,” commented Troy Margrie.

Their work is accessible in the journal Neuron. Another team, Lu et al. reporting in Nature, experimented on the ability to disambiguate self-motion from outside motion in arthropods. It requires changing coordinate systems from one reference frame to another — a common problem in calculus. The authors marvel at the ability of insects to navigate with behaviors that suggest complex vector calculations happening in their brains.

Insects can perform remarkable feats of navigation. For example, a desert ant can track its walking path using ‘dead reckoning’ (path integration), and the same is true of D. melanogaster. For accurate navigation, the brain needs to track the body’s velocity in all three degrees of freedom: rotation, forward translation and lateral translation…. Velocity information comes from sense organs—optic flow on the retina and mechanical input on limb proprioceptors—and probably also from copies of motor commands. Thus, velocity information arrives in body-centric coordinates. The brain must transform translational velocity signals into a world-centric coordinate frame by combining its estimate of body-centric translation direction… with its estimate of world-centric heading direction… to predict the animal’s world-centric travel direction.

Cassell discusses dead reckoning and four other navigation strategies in Animal Algorithms.

Navigating Abstract Space

A separate study at the University of Trento in Italy expanded neurological studies of navigation in real space to navigation in abstract space. They found that the ability to navigate unfamiliar spaces, even conceptual spaces, uses similar types of neural cells.

Some time ago, the research team provided experimental evidence that the populations of neurons that provide spatial information based on a grid-like system (as a navigation system) are activated not only to navigate the physical environment, but also to navigate in the abstract space of ideas and concepts. These neuronal populations are called place cells and grid cells, and are located in the hippocampus region and in the medial prefrontal cortex.

Now, the same research team — again using magnetic resonance imaging — has demonstrated that another type of brain cells is involved in abstract thinking, which is complementary to the previously identified types, and which acts as a sort of compass for orientation in the space of ideas.

Curiously, they found that navigating abstract space employs neurons that act like the head-direction cells:

In other words, they signal the direction of movement even when movement does not occur in a physical space, but in an abstract, conceptual space. This discovery suggests the existence of a complementary mechanism for conceptual navigation outside the hippocampal formation“, the research team concluded.

This work was published in the open-access journal Nature Communications Biology. Their experiments were conducted on humans only; they did not venture to speculate whether animals navigate conceptual space. The lack of philosophy textbooks by rats suggests that abstract spaces to them are limited to things like shapes, sounds, and other animals. We humans, however, are familiar with visualizing conceptual spaces and maneuvering through them with logic and reason, as if seeking the path of truth.

Hardware and Software 

It’s clear that navigational abilities of animals require both hardware and software. These recent papers reinforce the conclusions in Animal Algorithms by illustrating in more detail the irreducible complexity (IC) in both realms. The IC in hardware is not merely additive to the IC in software. Since both are interdependent, it’s like IC squared. If either realm is a challenge to Darwinian mechanisms, the product is vastly more challenging. Based on our uniform experience with explaining the origin of functional, information-rich systems, only intelligence is up to the task.