Week 3 Seminar: The trace

Memory without a nervous system, and the century-long mistake that comes with it — a finding discarded in 1906 because the replication used the wrong species.

Part II — Interior models11 March 2027Week 3 lectureSlidesSchedule

Physarum polycephalum has no neurons and solves the U-shaped trap problem, which is a standard test in robot navigation. It does this by avoiding its own slime trail. The memory is real and it is not inside the organism — it’s smeared on the substrate, and the organism reads it the way it reads anything else in the environment.

Reid and colleagues showed in 2012 that disguising the trail breaks the behaviour, which is what makes this a claim about memory rather than a story about mould. Note what the architecture buys: unlimited capacity, no maintenance cost, and total vulnerability to anyone with a pipette.

The part of this week that matters most#

In 1906 Jennings described a hierarchy of avoidance responses in the ciliate Stentor roeseli — bend away, alter the cilia, contract, detach — escalating under repeated stimulation. One cell, running something that looks like a decision procedure.

During the behaviourist era it was discredited as non-reproducible, and it stayed discredited for most of a century. Dexter, Prabakaran and Gunawardena replicated it in 2019 and found why the replications had failed: they used the wrong species. S. coeruleus is motile and builds no tube. S. roeseli is sessile. The two animals do not face the same problem, so of course they don’t behave the same way.

Ninety-three years, and the error is a species name. This is the shape of failure the course is actually about.

Sit with that, because it is not a story about fraud or about statistics. A correct finding was thrown away for ninety years because nobody checked a species name, and everyone involved was behaving reasonably. This is the failure mode this course is actually about, and it is much harder to guard against than dishonesty.

Two plant literatures, graded differently#

We finish on plants, where the evidence splits cleanly and instructively.

Gagliano and colleagues reported associative learning in pea plants in 2016. Markel failed to replicate it in 2020, using blinded scoring. Gagliano commented that the protocol was unsuitable; Markel responded that the differences could not account for the failure. Both exchanges are published, neither party conceded, and the matter is open. Grade: [C].

Dener, Kacelnik and Shemesh, also 2016, split pea root systems between a constant and a variable nutrient patch with the same mean. Plants grew more root into the variable patch when nutrients were low and less when they were high — risk-prone when poor, risk-averse when rich, which is the energy budget rule out of animal foraging theory. This one has held up. Grade: [E].

Two papers, same year, same plant, same broad claim family, and they do not get the same grade. You will be asked to explain why in the seminar, and “one replicated” is the beginning of the answer rather than the whole of it.

Reading#

The paper this seminar is built around:

• Dexter, Prabakaran and Gunawardena (2019). A complex hierarchy of avoidance behaviors in a single-cell eukaryote. Current Biology.

The rest of the week’s reading:

• Reid, Latty, Dussutour and Beekman (2012). Slime mould uses an externalized spatial memory to navigate in complex environments. PNAS.
• Nakagaki, Yamada and Tóth (2000). Maze-solving by an amoeboid organism. Nature.
• Gagliano and colleagues (2016). Learning by association in plants. Scientific Reports.
• Markel (2020). Lack of evidence for associative learning in pea plants. eLife — read this one directly after Gagliano, and read the published exchange that follows it.
• Dener, Kacelnik and Shemesh (2016). Pea plants show risk sensitivity. Current Biology — the plant result that did hold up.

Grades are not attached to papers here, because a grade belongs to a claim rather than to the paper it appears in — see evidence grades. Several of these papers contain claims at two different grades.