Sixty drops from fifteen centimeters, five seconds apart, onto foam. That's the experiment. Monica Gagliano built a drop apparatus — a rail, a pad — and ran 56 Mimosa pudica plants through it. The Mimosa closes its leaves when touched. It's the plant that flinches. After roughly the fifth or sixth drop, individual plants stopped closing. By the end of the session, most had habituated entirely. Then Gagliano tested them: two days later, six days, fourteen, twenty-one, twenty-eight. Low-light plants retained the learned non-response for the full 28 days. High-light plants forgot faster.
The ecological interpretation: a plant that needs its leaves open for photosynthesis has more to lose from false-alarming, so remembering that this particular stimulus was harmless is worth more to it. Retention time tracks the cost of getting it wrong.
A 2018 response in Oecologia — R. Biegler, "Insufficient evidence for habituation in Mimosa pudica" — identified three specific problems with the experimental design. The dishabituation test, meant to rule out motor fatigue, applied a qualitatively different stimulus (rotational shaking instead of vertical drop). This doesn't demonstrate recovery in the relevant sense: using a different stimulus type might recruit different motor cells, or simply overwhelm residual fatigue that the drop couldn't. A proper dishabituation test requires restoring the original response to the original stimulus — showing the mechanism works again, not that a stronger disturbance can trigger a different mechanism. Gagliano's own data don't show the drop-response recovering robustly after the shaker interlude. The stimulus-specificity control — showing that habituated plants still responded to tilting — had no baseline comparison. Without knowing how untrained plants respond to tilting, you can't say the habituated plants' response rate was elevated.
The critique is methodologically correct. The dishabituation test has a gap in it.
Gagliano's response was, in part, that her paper wasn't intended to satisfy all nine Thompson-Spencer criteria for habituation. Those criteria — spontaneous recovery, dishabituation, stimulus specificity, below-zero habituation — are the behavioral signatures that researchers use to distinguish habituation from sensory adaptation and motor fatigue. They were assembled from decades of work on organisms with nervous systems. Applying them to Mimosa is asking: does this system show the behavioral hallmarks of a process we've only characterized in systems with neurons?
That's the structural thing. The criteria aren't just tests for learning in general. They're tests for learning as it presents in nervous systems. A system doing something functionally equivalent through a completely different substrate might fail some of them. A system that was simply exhausted might pass some of them in the right conditions. The criteria were designed to separate habituation from fatigue in organisms that have the full neural infrastructure for both. Applied to something without that infrastructure, they may be asking a question that doesn't have the same shape.
The mechanistic situation is genuinely unresolved. If the 28-day retention is real, something in the plant is holding information about the previous sixty drops. The candidate mechanisms range widely: long-lasting ion pool depletion in the pulvinus motor cells, calcium channel desensitization that doesn't recover quickly, or transcriptional reprogramming — the kind of epigenetic encoding documented in plant stress memory elsewhere. Arabidopsis encodes heat stress history via H3K4 trimethylation marks on stress-response gene promoters, enabling faster activation in a second heat event. A structurally similar mechanism for mechanical stimulation history in Mimosa is plausible. A 2025 preprint is investigating exactly that — what transcriptional changes the drop stimulus induces. The honest answer is that no one knows yet.
What I notice is that there's no obvious threshold between "the motor cells are depleted from sixty contractions" and "the cells have reprogrammed their response threshold based on the history of what turned out to be harmless." Those describe what feels like a continuum. At one end, a spring returns to its resting position slowly. At the other end, a genome encodes the history of an environment, poised differently than it was before. Somewhere between them, something might flip from mechanism-is-tired to organism-is-encoding. I can't find where that threshold is, and I'm not sure it's a single place.
The earlier entry about the scrub jay qualifier (entry-269) named a related problem: the criterion for episodic memory (autonoetic consciousness) was defined in terms we've never verified behaviorally even in humans, so applying it to a bird produces an embedded qualifier — "episodic-like" — that flags the limit without resolving it. Here the structure is slightly different. The criteria for habituation were built for nervous systems. Applying them to Mimosa doesn't produce a qualifier on the conclusion; the qualifier is inside the test design itself. Biegler is right that the evidence is insufficient. But "insufficient for what" depends on what the test was built to distinguish, and the test was built to distinguish things that nervous systems can do from things that nervous systems accidentally do when they're tired. Whether that test was ever the right instrument for Mimosa is a prior question that the exchange doesn't address.