Skip to main content
NCBI home page
NIHPA Author Manuscripts logoLink to NIHPA Author Manuscripts
. Author manuscript; available in PMC: 2021 Apr 1.
Published in final edited form as: Neurobiol Learn Mem. 2019 Oct 11;170:107097. doi: 10.1016/j.nlm.2019.107097

Bidirectional short-term plasticity during single-trial learning of cerebellar-driven eyelid movements in mice

Farzaneh Najafi a, Javier F Medina b,*
PMCID: PMC7148184  NIHMSID: NIHMS1058670  PMID: 31610225

Abstract

The brain is constantly monitoring its own performance, using error signals to trigger mechanisms of plasticity that help improve future behavior. Indeed, adaptive changes in behavior have been observed after a single error trial in many learning tasks, including cerebellum-dependent eyeblink conditioning. Here, we demonstrate that the plasticity underlying single-trial learning during eyeblink conditioning in mice is bidirectionally regulated by positive and negative prediction errors, has an ephemeral effect on behavior (decays in<1 min), and can be triggered in the absence of errors in performance. We suggest that these three properties of single-trial learning may be particularly useful for driving mechanisms of motor adaptation that can achieve optimal performance in the face of environmental disturbances with a fast timescale.

Keywords: Cerebellum, Pavlovian conditioning, Motor learning, Eyeblink

1. Introduction

The role of the cerebellum in motor learning spans several timescales (Raymond & Medina, 2018). On the long end, the cerebellum is necessary for gradually improving our movements during hundreds of repetitions that may take place over the course of several hours in a training session, and also for consolidating the learned movements and retaining them for days, months or even years without further practice (Doyon & Benali, 2005; Krakauer & Shadmehr, 2006; Shadmehr & Holcomb, 1997). On the short end, the cerebellum has been implicated in single-trial learning (Diedrichsen, Verstynen, Lehman, & Ivry, 2005; Herzfeld, Kojima, Soetedjo, & Shadmehr, 2018; Junker, 2018; Kimpo, Rinaldi, Kim, Payne, & Raymond, 2014; Medina & Lisberger, 2008; Yang & Lisberger, 2014), a very fast type of adaptation that helps improve motor performance on the very next trial after an error has been committed. To understand how the cerebellum learns to optimize our movements in a changing world with both short-term and long-term environmental disturbances, it is important to define the properties of the learning processes that operate across these different timescales.

Eyeblink conditioning experiments are particularly well suited for investigating the mechanisms underlying cerebellum-dependent motor learning over multiple timescales. In this task, animals learn to make a protective blink in response to a cue (conditioned stimulus, CS) that is immediately followed by an aversive periocular shock or an airpuff directed at the eye (unconditioned stimulus, US). Previous studies have demonstrated that learning and memory consolidation of the eyeblink response are implemented in the cerebellum via mechanisms of plasticity that are slow and require multiple training sessions with hundreds of CS + US pairings each (Attwell, Cooke, & Yeo, 2002; Heiney, Wohl, Chettih, Ruffolo, & Medina, 2014; Medina, Garcia, & Mauk, 2001; Steinmetz & Freeman, 2016). However, a recent study has revealed that learning can also take place on much faster timescale, if errors are introduced after animals have been thoroughly trained (Khilkevich, Halverson, Canton-Josh, & Mauk, 2016): the protective eyeblink response was slightly smaller in trials that were preceded by an occasional presentation of the cue stimulus by itself, without the normal periocular stimulation (i.e. in trials immediately following a negative prediction error (Ohmae & Medina, 2015; Ludvig, Sutton, & Kehoe, 2012), NPE). This type of single-trial learning is thought to result from an NPE-driven error in motor performance (i.e. generating a CR that was not needed), which leads to a smaller CR in the next trial due to a decreased expectation that the cue will be followed by the aversive periocular stimulus immediately after the cue has been presented by itself.

To gain further insight into the mechanisms of plasticity underlying single-trial learning during eyeblink conditioning in mice, we performed three experiments designed to answer the following open questions: (1) is single-trial learning bidirectional, susceptible to positive prediction errors (PPE) in addition to negative prediction errors, (2) are performance errors necessary for single-trial learning, and (3) how long do motor adaptations resulting from single-trial learning last?

2. Methods

2.1. Animals

All experiments were performed in adult male C57BL/6 mice (The Jackson Laboratory). Mice were singly housed in a room with an inverted light/dark cycle, and all experiments took place during the dark phase. Procedures were performed in accordance with protocols approved by the University of Pennsylvania Animal Care and Use Committee based on guidelines of the National Institutes of Health.

2.2. Surgical procedures

Before all experiments, stereotaxic surgery was performed to implant a “head plate”. Mice were anesthetized with isoflurane (1.5–2% by volume in O2; SurgiVet) and kept on a heating pad to maintain body temperature. In addition, Meloxicam was given perioperatively to reduce swelling and to provide postoperative analgesia. A midline incision was made to expose the skull, and the underlying fascia was cleared with cotton swabs. Two small screws were inserted in the skull, on either side of the midline near bregma and a thin stainless steel head plate was then secured to the screws and skull using Metabond cement (Parkell).

2.3. Experimental design

2.3.1. Conditioning phase

All mice were habituated to head restraint for 2 days before beginning conditioning sessions, by placing them on top of a cylindrical treadmill inside a sound-isolated box and allowing them to walk at will while the head plate was attached to a pair of fixed rods. We then began conditioning sessions, comprising approximately 100 paired trials in which the onset of a conditioned stimulus (bright LED pulse, 270 ms in duration) was followed after 250 ms by an aversive unconditioned stimulus (20 psi airpuff of 20 ms duration, directed at the cornea via a 23 gauge needle placed 3 mm from the mouse’s eye). The interval of time between consecutive trials (ITI; intertrial interval) was set to 7 s. Mice were trained for > 20 conditioning sessions, until they had learned to protect the cornea from the aversive airpuff by making a conditioned eyelid response to the LED stimulus with high reliability (CRs made in > 80% of trials).

2.3.2. Triggers for plasticity underlying single-trial learning

For experiments in Figs. 1 and 2, we evaluated trial-to-trial learning by measuring CRs in test sessions that were identical to conditioning sessions except that the sixth trial out of every 12 trials was modified in one of the following ways: (1) In ‘No puff’ sessions the airpuff was omitted after the presentation of the LED stimulus, (2) In ‘Strong puff’ sessions the airpuff delivered after the LED stimulus was stronger than normal (40 psi, 40 ms), (3) In ‘Blank’ sessions the LED and the airpuff stimuli were both omitted, (4) In ‘Puff-alone’ sessions a stronger than normal airpuff (40 psi, 40 ms) was delivered without the preceding LED stimulus.

Fig. 1.

Fig. 1.

Open in a new tab

Bidirectional control of single-trial learning by prediction error. (A) Trial sequence during test sessions with occasional negative prediction error (NPE) trials in which the LED stimulus was presented without the airpuff (green). (B) Eyelid movements made by one mouse in one example session with large trial-to-trial changes after the NPE. Session-averaged eyelid traces are shown for all NPE trials (green), and for the previous (black) and the next trials (cyan). (C) The size of the conditioned response (CR), measured as the fraction of eyelid closure (FEC) in the time window indicated in (B), for all NPE trials from all mice (green), and all the previous (black) and next trials (cyan). (D) Average trial-to-trial changes in the eyelid movements, computed separately for each mouse by subtracting the FEC trace in every NPE trial from the FEC trace in the next trial (cyan), or in the previous trial (black). (E) Single-trial learning before (’control’) and after (’no-puff) an NPE trial, measured separately for each mouse by computing the average change in FEC in the CR window for the traces shown in (D). (F–J) Same as (A–E) but for test sessions with occasional positive prediction error (PPE) trials in which the LED stimulus was paired with a stronger-than-normal airpuff (green).

Fig. 2.

Fig. 2.

Open in a new tab

Non-associative contributions to single-trial learning. (A, B) Trial sequence during test sessions with occassional ‘blank’ trials in which both the LED and airpuff stimuli were omitted (A), or with occasional ‘puff-alone’ trials in which a stronger-than-normal airpuff stimulus was presented without the LED (B). (C) (Top) Average trial-to-trial changes in the eyelid movements, computed separately for each mouse by subtracting the FEC traces from pairs of trials, one before and one after the blank (cyan), or one before and two before the blank (black). (Bottom) Single-trial learning before (‘control’) and after (’blank’) a blank trial, measured separately for each mouse by computing the average change in FEC in the CR window for the traces shown in (C, Top). (D) Same as (C) but for test sessions with ‘puff-alone’ trials.

2.3.3. Timescale of plasticity underlying single-trial learning

For experiments in Fig. 3, we evaluated trial-to-trial learning by measuring CRs in the following test sessions: (1) Normal conditioning sessions except that the intertrial interval was varied in the range 3–80 s after the sixth trial out of every 12 trials, (2) Normal ‘No puff’ sessions except that the intertrial interval was varied in the range 3–80 s after trials in which the airpuff was omitted, (3) Normal ‘Strong puff’ sessions except that the intertrial interval was varied in the range 3–80 s after trials in which the strong airpuff was delivered.

Fig. 3.

Fig. 3.

Open in a new tab

Timescale of single-trial learning. (A–C) Trial sequence during test sessions with occasional NPE (A), PPE (B) or normal control (C) trials that were separated from the next trial by a variable intertrial interval (ITI) in the range 3–80 s. (D) Average trial-to-trial changes in the eyelid movements produced after a variable ITI, computed separately for each mouse by subtracting the FEC traces from the trial after and before the variable ITI in NPE sessions (cyan), PPE sessions (magenta) or control sessions (black). (E) Single-trial learning for different ITI values, computed as the difference in the CR window between the traces shown in (D) for the ‘control’ and the ‘strong-puff’ (magenta), or for the ‘control’ and ‘no-puff’ (cyan).

2.4. Data analysis

We measured the fraction of eyelid closure (FEC) as previously reported (Heiney et al., 2014), using videos of the eyelid movement recorded with a high-speed camera under infrared illumination (200 or 350 frames per second; Allied Vision). For each mouse, we used the FEC signal to compute two quantities: (1) ΔFEC(t), calculated as the average frame-by-frame subtraction of the FEC signal in N pairs of consecutive trials of a certain type, i.e. ΔFEC(t)=∑n=1N(FEC(t,trial(n+1))−FEC(t,trial(n)))/N and (2) %ΔCR size, calculated as the average ΔFEC(t) in the interval of time [240 250] ms after the onset of the LED stimulus, i.e. in the last 10 ms before the onset of the aversive puff, before any reflexive eyelid movements are generated.

3. Results

All the experiments reported below were performed after completion of a conditioning phase, in which head-fixed mice were thoroughly trained in our apparatus for eyeblink conditioning (Heiney et al., 2014) until they learned to make a protective eyelid conditioned response (CR) to the cue stimulus. This conditioning phase comprised > 20 training sessions with approximately 100 paired trials each. In every trial, an LED stimulus served as the cue and was presented 250 ms before the delivery of an aversive airpuff directed at the eye. All mice achieved high performance levels by the end of the conditioning phase (CRs made in > 80% of trials), indicating that they had learned to predict that the LED stimulus would be followed by the aversive airpuff with high reliability.

3.1. Single-trial learning is bidirectionally modulated by prediction error

Previous work during eyeblink conditioning in rabbits has demonstrated that the size of the CR is reduced on the trial immediately following a negative prediction error (Khilkevich et al., 2016) (NPE). The goal of our first experiment was to confirm this observation in the mouse, and to further characterize the properties of single-trial learning by assessing whether the size of the CR is also modified after a positive prediction error (PPE).

After completing the conditioning phase, mice (n = 8) received test sessions of normal paired trials that also included a small number of NPE trials (Fig. 1A–E, 8% trials with LED stimulus presented without the airpuff), or PPE trials (Fig. 1F–J, 8% trials with LED stimulus paired with an unusually strong airpuff). We observed that CRs in the trial immediately following an NPE were smaller than average (Fig. 1B and C; Wilcoxon signed rank test, P < 0.001), whereas CRs in the trial immediately following a PPE were bigger than average (Fig. 1G and H; Wilcoxon signed rank test, P < 0.001). To quantify the amount of single-trial learning for each mouse, we subtracted the FEC traces from pairs of consecutive trials and averaged together all the trial-to-trial changes in the eyelid movements generated by the mouse after an NPE (Fig. 1D, cyan; ‘no-puff’), before an NPE (Fig. 1D, black; ‘control’), after a PPE (Fig. 1I, magenta; ‘strong-puff’), or before a PPE (Fig. 1I, black; ‘control’). This analysis revealed that single-trial learning during eyeblink conditioning is bidirectionally modulated by prediction error, with NPE leading to a decrease in the size of the CR in the following trial (Fig. 1E; Wilcoxon signed rank test for n = 8 mice, P = 0.0078, mean difference between ‘no-puff’ and ‘control’ was −2.23%, 95% CI [−1.4, −3.4]), and PPE leading to an increase in the size of the CR in the following trial (Fig. 1J; Wilcoxon signed rank test for n = 8 mice, P = 0.0078, mean difference between ‘strong-puff’ and ‘control’ was +5.19%, 95% CI [+2.9, +7.2]). These changes in the size of the CR were the result of small differences in the frequency of trials with bigger and smaller than average CRs, and not to differences in the percentage of trials without a CR, i.e. trials with CR size<0.1 FEC (Supplementary Fig. 1).

3.2. Single-trial learning with and without errors in motor performance

There are two possible (and non-mutually exclusive) hypotheses about the source of the neural signal that triggers the mechanisms of plasticity underlying the single-trial changes in CR size observed in Fig. 1. First, the changes could be driven by an error in motor performance, i.e. a CR that is too big or too small in trials in which the strength of the aversive airpuff does not match what was expected to occur after the LED presentation. Second, the changes could be driven by signals related to environmental alterations in airpuff probability or strength, regardless of CR performance or whether these alterations were predicted by the LED stimulus or not. The goal of our second experiment was to distinguish between these two hypotheses, by assessing if LED presentations and CRs are necessary for single-trial learning.

We measured trial-to-trial changes in the CR performance of mice (n = 8), during test sessions that were identical to those in Fig. 1 except that the LED stimulus was omitted in all the occasional NPE trials (Fig. 2A, 8% ‘blank’ trials without LED or aipuff stimuli), and PPE trials (Fig. 2B, 8% ‘puff-alone’ trials with an unusually strong airpuff presented without a preceding LED stimulus). We observed a small decrease in the size of the CR after a blank trial (Fig. 2C, top; ‘blank’), which was not significantly different from the trial-to-trial change in CR size across the last two consecutive control trials before the blank (Fig. 2C, bottom; Wilcoxon signed rank test for n = 8 mice, P = 0.19, mean difference between ‘blank’ and ‘control’ was −1.38%, 95% CI [−2.7, +0.5]). In contrast, presentation of the airpuff by itself had a much more robust effect on the conditioned eyelid movement generated by the mice in the next trial (Fig. 2D, top; ‘puff-alone’), resulting in significant trial-to-trial increases in the size of the CR (Fig. 2D, bottom; Wilcoxon signed rank test for n = 8 mice, P = 0.0078, mean difference between ‘puff-alone’ and ‘control’ was +8.02%, 95% CI [+4.4, +10.1]). Thus, bidirectional single-trial learning can result from multiple processes, which can be triggered by NPE and PPE signals in trials with a CR (Fig. 1), as well as by contextual changes in stimulus parameters independent of CR performance (Fig. 2D).

3.3. Short-term persistence of single-trial learning

The mechanisms of plasticity underlying cerebellum-dependent motor learning are thought to operate over multiple timescales (Raymond & Medina, 2018), establishing memories that are used for different purposes and can persist for years or decay within seconds. The goal of our third experiment was to determine the rate of forgetting for the bidirectional single-trial learning of conditioned eyelid movements observed in Fig. 1.

Mice (n = 16) received test sessions that were identical to those in Fig. 1 except that the inter-trial time interval (ITI) between the occasional NPE or PPE trial and the next trial, which was kept at 7 s for the experiments in Fig. 1, varied randomly throughout the session over the range 3–80 s (Fig. 3A, ‘no-puff’ sessions; Fig. 3B, ‘strong-puff’ sessions). To control for any ITI-dependent decay in the size of the CR that may occur during normal conditioning without NPE or PPE trials, mice also received sessions in which the ITI varied occasionally over the same 3–80 s range, but the strength of the airpuff was kept constant in every trial (Fig. 3C, ‘control’ sessions).

The results shown in Fig. 3D and E reveal three key features about the different ways in which the passage of time impacted conditioned eyelid movements in our experiments: (1) The size of the CR decayed gradually after a normal paired trial in control sessions (Fig. 3D, ‘control’), (2) The size of the CR was different from controls in the first 10 s and up to 30 s after a prediction error, with PPE leading to bigger CRs (Fig. 3D and E, Table 1; ‘strong-puff vs ‘control’, Dunnett’s test, P < 0.001 for ITI 10–15, 15–30; P < 0.05 for ITI 30–60) and NPE leading to smaller CRs (Fig. 3D and E, Table 1; ‘no-puff’ vs ‘control’, Dunnett’s test, P < 0.01 for ITI 3–10, 10–15; P < 0.05 for ITI 15–30). (3) The size of the CR was not different from controls 60–80 s after a NPE or PPE trial (Fig. 3D and E, Table 1; Dunnett’s test, P = 0.73 for ‘strong-puff’ vs ‘control; P = 0.97 for ‘no-puff’ vs ‘control’). These results suggest that the plasticity underlying single-trial learning during eyeblink conditioning has fast on/off kinetics, i.e. it is induced in < 10 s and then decays quickly with a time constant < 1 min. A similar time constant has been reported for single-trial learning in cerebellar-driven smooth pursuit eye movements in monkeys (Yang & Lisberger, 2010).

Table 1.

Descriptive statistics for Dunnett’s multiple comparisons test in Fig. 3.

Comparison ITI range Mean diff 95% CI p-value
‘No puff’ vs ‘Control’ 3–10 −4.78 [−8.51, −1.05] 0.009
‘No puff’ vs ‘Control’ 10–15 −5.80 [−9.45, −2.15] 0.001
‘No puff’ vs ‘Control’ 15–30 −4.57 [−8.22, −0.92] 0.011
‘No puff’ vs ‘Control’ 30–60 −2.33 [−5.99, +1.32] 0.265
‘No puff’ vs ‘Control 60–80 +0.31 [−3.34, +3.97] 0.973
‘Strong puff’ vs ‘Control’ 3–10 +3.42 [−0.23, +7.08] 0.070
‘Strong puff’ vs ‘Control’ 10–15 +6.69 [+3.04, +10.35] 0.0001
‘Strong puff’ vs ‘Control’ 15–30 +6.07 [+2.42, +9.73] 0.0006
‘Strong puff’ vs ‘Control’ 30–60 +4.44 [+0.78, +8.10] 0.014
‘Strong puff’ vs ‘Control’ 60–80 +0.32 [−2.57, +4.74] 0.733
Open in a new tab

4. Discussion

4.1. Summary of results

Our results reveal three properties of single-trial learning, as measured during performance of conditioned eyelid responses (CRs) in the mouse: (1) Single-trial learning is bidirectional and can lead to either bigger or smaller CRs depending on stimulus parameters, (2) There are multiple triggers for single-trial learning, including a negative prediction error (NPE, unexpected omission of the airpuff after presentation of the LED stimulus), a positive prediction error (PPE, stronger than normal airpuff after presentation of the LED stimulus), or a sensitizing stimulus (SS, stronger than normal airpuff delivered by itself), (3) The effects of single-trial learning on CR size are transient, persisting for approximately 60–80 s. These findings provide some important clues about the functional role of single-trial learning during cerebellar-driven behavior, and about the underlying mechanisms of plasticity, both of which are discussed below.

4.2. Neural plasticity underlying single-trial learning

Previous studies have identified the key brain regions responsible for supporting CR performance during eyeblink conditioning (Freeman & Steinmetz, 2011). In the rodent, the critical nodes include Purkinje cells located at the bottom of the primary fissure in the cerebellar cortex (Heiney, Kim, Augustine, & Medina, 2014; Ohmae & Medina, 2015; Steinmetz & Freeman, 2014; ten Brinke, 2015), and neurons in the anterior interpositus of the cerebellar nuclei (Heiney et al., 2014; Ten Brinke, 2017). In theory, a single-trial change in the size of the conditioned eyelid response to the LED stimulus could be implemented by mechanisms of plasticity that modulate activity in the upstream sensory pathways that send information about the LED to the cerebellum (Freeman & Steinmetz, 2011), in the cerebellum itself (Freeman & Steinmetz, 2011), or in the downstream motor circuits that produce a blink (Freeman & Steinmetz, 2011). Based on recent work demonstrating that neural correlates of single-trial learning are present at the level of Purkinje cells (Herzfeld et al., 2018; Kimpo et al., 2014; Medina & Lisberger, 2008; Yang & Lisberger, 2014), and that single-trial learning still occurs when the LED stimulus is substituted with direct stimulation of mossy fiber inputs (Khilkevich et al., 2016), we favor the hypothesis of a plasticity site within the cerebellar cortex.

Our behavioral results suggest that the mechanisms of plasticity supporting single-trial learning during eyeblink conditioning should be bidirectional, short-lived, and sensitive to instructive signals whose efficacy can be regulated by the predictability and the strength of an airpuff stimulus. Within the cerebellar cortex, a candidate mechanism for the single-trial changes in CR size observed after PPE and NPE events is short-term plasticity of LED-activated synapses of parallel fibers onto Purkinje cells (Brenowitz & Regehr, 2005; Goto, Inoue, Kuruma, & Mikoshiba, 2006), which is under the control of sensory-driven climbing fiber inputs whose strength is graded and depends on airpuff parameters like its intensity (Gaffield, Bonnan, & Christie, 2019; Najafi, Giovannucci, Wang, & Medina, 2014a, 2014b), and the context in which it is delivered (Carey & Regehr, 2009; Najafi & Medina, 2013; Ohmae & Medina, 2015; Rasmussen, 2019). In contrast, the single-trial learning observed after an SS event requires a short-term mechanism of heterosynaptic and non-associative plasticity that is able to change how the mouse will respond to the LED stimulus after the aversive airpuff is delivered by itself (without the LED) (Bailey, Giustetto, Huang, Hawkins, & Kandel, 2000). Although it is well known that there are multiple sites of short-term plasticity in the neural pathways supporting CR performance (Bracha & Bloedel, 1996), including many inside the cerebellum (Carey, 2011; Gao, van Beugen, & De Zeeuw, 2012), more work will be necessary to understand how (if) they contribute to single-trial learning during eyeblink conditioning.

4.3. Functional implications

The properties of single-trial learning uncovered in our experiments appear to be ideally suited for driving mechanisms of motor adaptation that can achieve optimal performance in the face of environmental disturbances with a fast timescale (Kording, Tenenbaum, & Shadmehr, 2007; Wei & Kording, 2009). For example, if the tennis ball goes out on the first serve because a strong tail wind suddenly starts blowing and pushes it too far (i.e. a prediction error), single-trial learning would help achieve better results on the second serve by momentarily reducing the magnitude of the motor command. Short-term plasticity is useful in this scenario because wind strength and direction are temporary disturbances, and making permanent modifications to the motor command would be maladaptive and lead to poor performance in the long run. In addition, we have found that single-trial learning can be triggered without errors in motor performance (i.e. after a strong airpuff was delivered by itself), a feature of the underlying mechanisms of plasticity that may be particularly useful for adapting motor commands when feedback information about performance errors is not available, e.g. the wind started blowing just before the first serve.

Cerebellar-dependent motor learning recruits multiple sites of plasticity that work together to adaptively modify behavior over a wide range of timescales (Boyden, Katoh, & Raymond, 2004; Medina, Nores, Ohyama, & Mauk, 2000; Raymond & Medina, 2018; Yang & Lisberger, 2014). Previous work has suggested that single-trial learning leads to, and may even be a necessary prerequisite for, some components of long-term learning (Khilkevich et al., 2016; Smith, Ghazizadeh, & Shadmehr, 2006; Yang & Lisberger, 2014), although there is also evidence that the underlying mechanisms of plasticity for learning at these two extreme timescales can be partially independent of each other (Kimpo et al., 2014; Yang & Lisberger, 2014). Our experiments cannot speak to this issue because they were designed to assess single-trial learning after mice had been thoroughly trained and conditioned movements had been firmly established. Understanding the links between single-trial learning and long-term modification of motor performance could help improve the efficacy of rehabilitation protocols and is an important question for the future.

Supplementary Material

Supplementary Figure 1

Acknowledgments

Supported by grants R01MH093727 and RF1MH114269 to JM from the National Institutes of Health.

Footnotes

Appendix A. Supplementary material

Supplementary data to this article can be found online at https://doi.org/10.1016/j.nlm.2019.107097.

Declaration of Competing Interest

The authors declare no competing financial interests.

References

  1. Attwell PJ, Cooke SF, & Yeo CH (2002). Cerebellar function in consolidation of a motor memory. Neuron, 34, 1011–1020. [DOI] [PubMed] [Google Scholar]
  2. Bailey CH, Giustetto M, Huang YY, Hawkins RD, & Kandel ER (2000). Is heterosynaptic modulation essential for stabilizing Hebbian plasticity and memory? Nature Reviews Neuroscience, 1, 11–20. [DOI] [PubMed] [Google Scholar]
  3. Boyden ES, Katoh A, & Raymond JL (2004). Cerebellum-dependent learning: The role of multiple plasticity mechanisms. Annual Review of Neuroscience, 27, 581–609. [DOI] [PubMed] [Google Scholar]
  4. Bracha V, & Bloedel JR (1996). The acquisition of motor behavior in vertebrates (pp. 175–204). Cambridge, MA: MIT Press. [Google Scholar]
  5. Brenowitz SD, & Regehr WG (2005). Associative short-term synaptic plasticity mediated by endocannabinoids. Neuron, 45, 419–431. [DOI] [PubMed] [Google Scholar]
  6. Carey MR (2011). Synaptic mechanisms of sensorimotor learning in the cerebellum. Current Opinion in Neurobiology, 21, 609–615. [DOI] [PubMed] [Google Scholar]
  7. Carey MR, & Regehr WG (2009). Noradrenergic control of associative synaptic plasticity by selective modulation of instructive signals. Neuron, 62, 112–122. [DOI] [PMC free article] [PubMed] [Google Scholar]
  8. Diedrichsen J, Verstynen T, Lehman SL, & Ivry RB (2005). Cerebellar involvement in anticipating the consequences of self-produced actions during bimanual movements. Journal of Neurophysiology, 93, 801–812. [DOI] [PubMed] [Google Scholar]
  9. Doyon J, & Benali H. (2005). Reorganization and plasticity in the adult brain during learning of motor skills. Current Opinion in Neurobiology, 15, 161–167. [DOI] [PubMed] [Google Scholar]
  10. Freeman JH, & Steinmetz AB (2011). Neural circuitry and plasticity mechanisms underlying delay eyeblink conditioning. Learning and Memory, 18, 666–677. [DOI] [PMC free article] [PubMed] [Google Scholar]
  11. Gaffield MA, Bonnan A, & Christie JM (2019). Conversion of graded presynaptic climbing fiber activity into graded postsynaptic Ca(2+) signals by purkinje cell dendrites. Neuron. [DOI] [PMC free article] [PubMed] [Google Scholar]
  12. Gao Z, van Beugen BJ, & De Zeeuw CI (2012). Distributed synergistic plasticity and cerebellar learning. Nature Reviews Neuroscience, 13, 619–635. [DOI] [PubMed] [Google Scholar]
  13. Goto J, Inoue T, Kuruma A, & Mikoshiba K. (2006). Short-term potentiation at the parallel fiber-Purkinje cell synapse. Neuroscience Research, 55, 28–33. [DOI] [PubMed] [Google Scholar]
  14. Heiney SA, Kim J, Augustine GJ, & Medina JF (2014). Precise control ofmovement kinematics by optogenetic inhibition of Purkinje cell activity. Journal of Neuroscience, 34, 2321–2330. [DOI] [PMC free article] [PubMed] [Google Scholar]
  15. Heiney SA, Wohl MP, Chettih SN, Ruffolo LI, & Medina JF (2014). Cerebellar-dependent expression of motor learning during eyeblink conditioning in head-fixed mice. Journal of Neuroscience, 34, 14845–14853. [DOI] [PMC free article] [PubMed] [Google Scholar]
  16. Herzfeld DJ, Kojima Y, Soetedjo R, & Shadmehr R. (2018). Encoding of error and learning to correct that error by the Purkinje cells of the cerebellum. Nature Neuroscience, 21, 736–743. [DOI] [PMC free article] [PubMed] [Google Scholar]
  17. Junker M, et al. (2018). Learning from the past: A reverberation of past errors in the cerebellar climbing fiber signal. PLoS Biology, 16, e2004344. [DOI] [PMC free article] [PubMed] [Google Scholar]
  18. Khilkevich A, Halverson HE, Canton-Josh JE, & Mauk MD (2016). Links between single-trial changes and learning rate in eyelid conditioning. Cerebellum, 15,112–121. [DOI] [PubMed] [Google Scholar]
  19. Kimpo RR, Rinaldi JM, Kim CK, Payne HL, & Raymond JL (2014). Gating of neural error signals during motor learning. Elife, 3, e02076. [DOI] [PMC free article] [PubMed] [Google Scholar]
  20. Kording KP, Tenenbaum JB, & Shadmehr R. (2007). The dynamics of memory as a consequence of optimal adaptation to a changing body. Nature Neuroscience, 10, 779–786. [DOI] [PMC free article] [PubMed] [Google Scholar]
  21. Krakauer JW, & Shadmehr R. (2006). Consolidation of motor memory. Trends in Neurosciences, 29, 58–64. [DOI] [PMC free article] [PubMed] [Google Scholar]
  22. Ludvig EA, Sutton RS, & Kehoe EJ (2012). Evaluating the TD model of classical conditioning. Learning and Behavior, 40, 305–319. [DOI] [PubMed] [Google Scholar]
  23. Medina JF, Garcia KS, & Mauk MD (2001). A mechanism for savings in the cerebellum. Journal of Neuroscience, 21, 4081–4089. [DOI] [PMC free article] [PubMed] [Google Scholar]
  24. Medina JF, & Lisberger SG (2008). Links from complex spikes to local plasticity and motor learning in the cerebellum of awake-behaving monkeys. Nature Neuroscience, 11, 1185–1192. [DOI] [PMC free article] [PubMed] [Google Scholar]
  25. Medina JF, Nores WL, Ohyama T, & Mauk MD (2000). Mechanisms of cerebellar learning suggested by eyelid conditioning. Current Opinion in Neurobiology, 10, 717–724. [DOI] [PubMed] [Google Scholar]
  26. Najafi F, Giovannucci A, Wang SS, & Medina JF (2014a). Coding of stimulus strength via analog calcium signals in Purkinje cell dendrites of awake mice. Elife, 3, e03663. [DOI] [PMC free article] [PubMed] [Google Scholar]
  27. Najafi F, Giovannucci A, Wang SS-H, & Medina JF (2014b). Sensory-driven enhancement of calcium signals in individual purkinje cell dendrites of awake mice. Cell Reports, 6, 792–798. [DOI] [PMC free article] [PubMed] [Google Scholar]
  28. Najafi F, & Medina JF (2013). Beyond “all-or-nothing” climbing fibers: Graded representation of teaching signals in Purkinje cells. Frontiers in Neural Circuits, 7, 115. [DOI] [PMC free article] [PubMed] [Google Scholar]
  29. Ohmae S, & Medina JF (2015). Climbing fibers encode a temporal-difference prediction error during cerebellar learning in mice. Nature Neuroscience, 18, 1798–1803. [DOI] [PMC free article] [PubMed] [Google Scholar]
  30. Rasmussen A. (2019). Graded error signals in eyeblink conditioning. Neurobiology of Learning and Memory. [DOI] [PubMed] [Google Scholar]
  31. Raymond JL, & Medina JF (2018). Computational principles of supervised learning in the cerebellum. Annual Review of Neuroscience, 41, 233–253. [DOI] [PMC free article] [PubMed] [Google Scholar]
  32. Shadmehr R, & Holcomb HH (1997). Neural correlates of motor memory consolidation. Science, 277, 821–825. [DOI] [PubMed] [Google Scholar]
  33. Smith MA, Ghazizadeh A, & Shadmehr R. (2006). Interacting adaptive processes with different timescales underlie short-term motor learning. PLoS Biology, 4, e179. [DOI] [PMC free article] [PubMed] [Google Scholar]
  34. Steinmetz AB, & Freeman JH (2014). Localization of the cerebellar cortical zone mediating acquisition of eyeblink conditioning in rats. Neurobiology of Learning and Memory, 114, 148–154. [DOI] [PMC free article] [PubMed] [Google Scholar]
  35. Steinmetz AB, & Freeman JH (2016). Cannabinoid modulation of memory consolidation within the cerebellum. Neurobiology of Learning and Memory, 136, 228–235. [DOI] [PMC free article] [PubMed] [Google Scholar]
  36. ten Brinke MM, et al. (2015). Evolving models of pavlovian conditioning: Cerebellar cortical dynamics in awake behaving mice. Cell Reports, 13, 1977–1988. [DOI] [PMC free article] [PubMed] [Google Scholar]
  37. Ten Brinke MM, et al. (2017). Dynamic modulation of activity in cerebellar nuclei neurons during pavlovian eyeblink conditioning in mice. Elife, 6. [DOI] [PMC free article] [PubMed] [Google Scholar]
  38. Wei K, & Kording K. (2009). Relevance of error: What drives motor adaptation? Journal of Neurophysiology, 101, 655–664. [DOI] [PMC free article] [PubMed] [Google Scholar]
  39. Yang Y, & Lisberger SG (2010). Learning on multiple timescales in smooth pursuit eye movements. Journal of Neurophysiology, 104, 2850–2862. [DOI] [PMC free article] [PubMed] [Google Scholar]
  40. Yang Y, & Lisberger SG (2014). Role of plasticity at different sites across the time course of cerebellar motor learning. Journal of Neuroscience, 34, 7077–7090. [DOI] [PMC free article] [PubMed] [Google Scholar]
  41. Yang Y, & Lisberger SG (2014). Purkinje-cell plasticity and cerebellar motor learning are graded by complex-spike duration. Nature. [DOI] [PMC free article] [PubMed] [Google Scholar]

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

Supplementary Figure 1
NIHMS1058670-supplement-Supplementary_Figure_1.pdf (231KB, pdf)

RESOURCES