Article Figures & Data
Figures
- Figure 1.
Approaches used to study learning in the direction of smooth pursuit eye movements. A, From top to bottom, the zigzags indicate the trajectories of the learning target motions in the repeated-direction, random-order, and alternating paradigms. B, The superimposed traces show horizontal and vertical velocity as a function of time from the onset of target motion in example trials from one learning block. Dashed and solid traces show target and eye movement. Black and blue traces show responses in the first versus 100th off-direction learning trials in the repeated-direction paradigm. The red and black arrowheads on the velocity records point out the onset of the initial pursuit and the peak of the learned response in eye movements. The gray shading shows the analysis interval for learning. C, Red and black traces show the trial-over-trial change in firing rate versus time for pairs of trials with versus without a CS in the instruction trial. Vertical dashed line shows the time of the instructive change in target direction. C is reprinted with permission from Yang and Lisberger (2013). D, Simple-spike firing and CS of a representative Purkinje cell in an off-direction learning trial. The red asterisk indicates a CS. E, Raster shows the occurrence of CS responses in relation to the time of the instruction. Black curve shows the probability of CS responses in 100 ms bins. Gray shading shows the analysis interval for CS responses. F, Different color traces show the trial-over-trial change in firing rate versus time for pairs of trials with different durations of CSs in the instruction trial. Vertical dashed line shows the time of the instruction. F is reprinted with permission from Yang and Lisberger (2014a).
- Figure 2.
Trial-over-trial learning and properties of CS responses to the instruction with different sequences of learning tasks. A, C, D, Time course of trial-over-trial changes in eye velocity in the analysis interval for learning in blocks of 10 trials. Different graphs show data for repeated (A), alternating (C), and random-direction (D) paradigms. B, Time course of absolute eye velocity in the analysis interval for learning in the repeated paradigm, again in blocks of 10 trials. E, F, The probability (E) and duration (F) of a CS response to an instruction as a function of learning paradigm. The sets of three connected symbols present data for different Purkinje cells tested in all three learning paradigms. Ra, random; Re, repeated; Al, alternating paradigm. After Bonferroni correction: p(Ra, Re) = 1.75 × 10–12 and p(Ra, Al) = 1.1 × 10–10 for probability; p(Ra, Re) = 1.45 × 10–3 and p(Ra, Al) = 0.02 for duration.
- Figure 3.
Time courses of pursuit learning in eye movements, trial-over-trial depression of simple-spike firing, CS probability, and CS duration when instructions are present in repeated, random-order, or alternating sequences. From left to right, the three columns summarize the data for the repeated, random, and alternative sequences. Each graph shows a learning curve for data analyzed in bins of 10 trials. A–C, Trial-over-trial changes in eye velocity. D–F, Trial-over-trial changes in simple-spike firing of Purkinje cells. G–I, Probability of CS responses. J–L, Duration of CS responses. Error bars: ± 1SE of the mean. Blue symbols in right column repeat those in left column.
- Figure 4.
Controls showing that effect of learning paradigm depends on broader context of sequence of learning instructions. A, Experimental design for delivering instructions with target speeds of 29, 30, or 31 deg/s. B, CS probability versus time for target motion at different speeds. In A and B, traces of different colors show data for different target speeds. C, D, CS probability as a function of the history of direction of instruction for the prior four trials. In C, we selected from the random paradigm sequences of trials that repeat the off-direction target motion for up to four trials. Bar labeled Re shows CS probability in the repeated paradigm, and bars labeled with numbers show CS probability in the random-direction paradigm after sequences of one, two, three, or four instructions of the same direction. In D, we performed the same analysis for sequences of alternating direction instructions during the random direction paradigm.
- Figure 5.
Comparison of time courses of simple-spike learning in the on- and off-directions for the repeated, random, and alternating-direction paradigms. Each symbol plots the simple-spike firing in the interval from 100 ms before to 50 ms after the onset of an instructive change in target direction, with the baseline simple-spike firing rate subtracted. Filled and open symbols show the firing rate in trials that followed and on-direction versus and off-direction instruction. Error bars show SEMs across the 34 Purkinje cells in the sample. Gray shading indicates the first 50 trials in each block.
- Figure 6.
Low correlation between CS probability and duration measured in bins of 10 trials. A, Graph plots CS duration versus probability. Open and filled symbols show data from two Purkinje cells, and each symbol shows measurements for a single bin of 10 trials. Lines were obtained by regression analysis. B, Distribution of correlation between probability and duration of CSs. Blue bars show data from our sample of Purkinje cells. Red and green traces show predictions of the model described later in the paper, when the underlying correlation between CS probability was 1 (red) or 0.5 (green) and the degree of temporal decorrelation was set at 0.8. C, Full grid of measured correlation between CS duration and probability in 10-trial bins as a function of the parameters of the model: underlying correlation between duration and probability and temporal correlation in CS duration value (temporal correlation in this graph is equal to 1 minus the temporal decorrelation parameter of the model). Black curve shows the values that yielded a measured correlation between CS duration and probability equal to 0.1 and defines the upper limit of biologically compatible values of the model parameters. Red traces outline the region that is compatible with the measured temporal correlation in CS duration and delimits the pixels below that are compatible with all the data.
- Figure 7.
Absence of temporal correlation in duration of CS responses. A, Duration of all CS responses in off-direction learning trials during a random direction learning block for three representative Purkinje cells. B, Distribution of correlations between durations of CS responses in pairs of successive off-direction learning trials. C, Distribution of correlations between durations of CS responses in randomly chosen pairs of trials. Number of observations is divided by 10, because we performed this analysis 10 times for each Purkinje cell, selecting the same number of pairs of trials as we had pairs of consecutive CS responses for that cell. The red curve in B and C reproduces the control distribution in C.
- Figure 8.
Agreement between statistics of model and data. A, Structure of the relationship between partial correlations trial-over-trial changes in simple-spike firing with CS probability on the x-axis and duration on the y-axis. Symbols show different neurons. B–D, The same analysis as in A, but for the computational model with the underlying correlation of duration and probability set to 0.2, 0.5, or 0.8 and the temporal decorrelation set to 0.8. Blue and red filled symbols show the averages for the data and the model, respectively. E, F, Distributions of trial-over-trial changes in simple-spike firing, contingent on the presence (E) or absence (F) of a CS response to the instructive change in target direction on the first trial of the pair. Black histograms show the data across all Purkinje cells, and the red curve shows the best-fitting normal distribution, also used in the model. G, Distribution of CS probability. Black histogram shows analysis of the data from all Purkinje cells in bins of 10 trials, and the red dashed curve shows the same analysis of the model Purkinje cells. H, Distribution of CS durations. Histogram bars show the data based on single trials, continuous blue and dashed red curves show results for the data and the model, based on averaging across bins of 10 trials.
In this issue
