Experimental setup for studying single-cell-level habituation and potentiation in Stentor.

(A) Experimental apparatus: 20–30 Stentor cells are placed in a dish above a solenoid that delivers taps. Overhead video recordings are analyzed to extract contraction events. (B) Image showing a single Stentor extended before a tap and contracted after the tap. (C) The experimental protocol consisted of a first trial with a sequence of mechanical stimuli at a fixed inter-stimulus interval (ISI), followed by an inter-trial interval (ITI), followed by a second trial identical to the first. This yields a set of binary contraction responses for each cell at each stimulus during each trial. (D) Population mean response as a proxy for the latent contraction probability. Cells habituate to the series of stimuli, recover close to their baseline response level after a recovery period, and habituate faster in the second trial compared to the first.

Statistical model recovers naive analysis of response probability.

(A) Schematic showing the parametrization of the Hill function. Each parameter controls a corresponding aspect of curve shape. (B) Each row shows the inferred latent response probability for each cell at each stimulus. Cells have been sorted by cumulative response probability across both trials. (C) We obtain a unique latent probability function for each cell that can be averaged to get a point estimate at the population level. (D) The mean of the single-cell response curves faithfully recovers the population mean. (E) Single-cell curves, plotted with 95% credible interval bands, decrease much more sharply than the population mean would suggest. The population mean decreases more gradually due to heterogeneity in half-max times.

Quantifying relationships between response-profile parameters across cells.

(A) The posterior for the Hill coefficient is centered far above 1, supporting switch-like habituation at the single-cell level for this 1 min ISI condition. (B) Distribution of N50 values across cells for the two trials. (C) Distribution of initial response ratios across cells, showing heterogeneous recovery. (D) Most cells have a lower N50 in the second trial, indicated by a ratio below 1. (E–H) Scatter plots showing pairwise relationships between inferred parameters across cells. Cells are binned into quintiles by the x-axis parameter; points show within-bin medians; error bars show the central 95% within-bin quantile range of the cell-level posterior median estimates. (E) Relationship between trial-1 initial response probability and trial-2 N50. (F) Relationship between trial-1 initial response probability and recovery ratio. (G) Relationship between trial-1 N50 and N50 ratio. (H) Relationship between recovery ratio and N50 ratio.

Frequency sensitivity of Stentor habituation.

(A) We varied the ISI from 1 min to 3 min. As a function of stimulus number, the population mean response looks similar across ISIs. (B) Inferred trial-1 response curves plotted as a function of stimulus number. (C) The same inferred curves plotted as a function of time; shorter ISI leads to faster habituation in absolute time. (D) The Hill coefficient is higher for the 1 min ISI condition than for the 2 or 3 min conditions, suggesting a switchier habituation profile at the single-cell level. Error bars show 95% credible intervals. (E) Median half-max values are similar across ISIs when measured in stimulus-number units. (F) Recovery is frequency-sensitive: recovery ratios after a 1 hr ITI decrease with ISI.

Inferred single-cell curves plotted with population parameters.

Each axis displays the inferred single-cell curves for a condition with a particular ISI (1, 2, 3 minutes) and ITI (1, 2, 3, 5 hours). Bands show curves corresponding to the 95% credible interval.

Inferred phase portraits.

Each axis displays the phase portraits computed using the inferred single-cell curves for a condition with a particular ISI (1, 2, 3 minutes) and ITI (1, 2, 3, 5 hours). Scatter points show progression by stimulus number.

Quantifying potentiation across two trials.

(A) Ratio between N50 (half-max) values. (B) Ratio between initial learning rates, discretized by averaging across the first five stimuli. (C) Cumulative distance between phase portraits captures deviation in dynamics.

Quantifying potentiation across ITIs.

(A) Initial response recovery curves for each ISI condition across all recovery periods with 95% credible intervals. (B) Log N50 advantage decreases toward zero with increasing ITI. (C) Initial learning-rate ratio similarly decays with rest duration. (D) Cumulative phase-portrait difference decays with rest duration. (E–G) Recovery (x-axis) plotted against each potentiation metric (y-axis) for all ISI/ITI combinations. The spread across ISI conditions at similar recovery levels suggests partial decoupling of the recovery and potentiation processes.