Stir
Mimosa pudica already has one room here, Touch — a leaflet folding shut in a fraction of a second under a finger. This is the same plant doing something much slower and much stranger: nothing touches it at all, and hours before its light turns on, it starts moving anyway.
Vishton & Bartosh (Cognitive Science, December 2025) ran Mimosa pudica seedlings through a repeating three-day cycle — two days of 12 hours light and 12 hours dark, then a full day of nothing but dark — for 74 days, filming the whole time and measuring leaf and stem motion frame by frame. After about five repetitions of the cycle, a pattern showed up in the dark hours right before light was due: more movement on the two days light was actually coming, none of that rise on the all-dark third day — even though, at the moment measured, the plants had already been sitting in identical darkness for hours on either kind of day. Something in there could tell the difference. Then the authors changed the day length itself, compressing the whole cycle from 24 hours to 20, and the anticipatory rise showed up again on the very first compressed cycle — no days of readjusting. A plant running on an internal clock built for 24-hour days should have needed time to catch up. These didn't, which is the authors' own case for calling this counting discrete events rather than simply tracking elapsed time.
Below is a simplified version of that same design. Advance the clock hour by hour (or let it run) through the light–light–dark cycle, and watch the bars: dark hours close to an oncoming light period get taller as more cycles complete, dark hours on the all-dark day don't. Then try compressing the day length without resetting anything, the way the real experiment did.
Read the chart as a rolling window on the last several days. Pale bars are light hours; dark bars are dark hours, and their height is how much the simulated plant is moving. Early on, every dark hour looks about the same — low, flat, a little random. By the fifth cycle or so, watch the dark hours right before a light day's own dawn start rising above that baseline, while the all-dark third day stays flat almost all the way through. Switch the day length once that pattern is established and it carries straight over — the plant doesn't start over from a flat baseline the way it did on cycle one.
Honest gap: this room models anticipation as a direct function of simulated hours-until-next-light, rising inside a fixed six-hour pre-light window and scaled by a training level that ramps from 0 to full over five completed cycles — the paper's own reported timeline for when the pattern first appeared, not a curve the paper itself plots. The real result compared movement in a fixed pre-dawn clock window (roughly midnight to 6 a.m.) across day types; this room compares by hours-until-light instead, which is a cleaner match for what "counting" would actually predict but isn't the exact measurement the authors made. One consequence worth naming: under this room's own model, the all-dark day's final hours also rise, because light really is only hours away by then — a plausible extrapolation of the paper's own logic, not something it measured or claims. The "random 10–32 hour" phase adds extra noise and unreliability on top of the trained signal, standing in for the paper's own hedges about that condition — "higher levels of variability," results that "must be interpreted cautiously," day-length boundaries that were "post hoc" rather than pre-registered — not a plotted uncertainty band. The real study never analyzed individual plants at all: 18 cups of five seeds each were measured as one pooled signal, and the authors flag that as open ground for future work, not a settled finding about any one plant. This room's single leaf stands in for that whole pooled measurement, the same liberty the paper's own aggregate analysis already takes. And the real measurement — pixel change in time-lapse video — doesn't distinguish which structures moved or why; this room's leaf glyph wiggling is illustration, not a claim about which hinge is doing the anticipating.
Source: Vishton, P.M. & Bartosh, P.J., Can Mimosa pudica Plants Enumerate Light Exposure Events?, Cognitive Science 49(12), e70161 (2025), doi:10.1111/cogs.70161.