Encoding by Response Duration in the Basal Ganglia

Naama Parush, David Arkadir, Alon Nevet, Genela Morris, Naftali Tishby, Israel Nelken, Hagai BergmanJournal of Neurophysiology 2008 · DOI

TL;DR

This paper reveals that the duration of neuronal responses in the basal ganglia, not just their firing rate, carries significant information about abstract cues predicting future reward. This finding supports "gating" models of basal ganglia function and suggests that response duration is a crucial, underappreciated feature for information transmission within this brain region.

What Problem Does This Paper Address?

The basal ganglia (BG) are a group of subcortical nuclei involved in motor control, learning, and decision-making. A central question in understanding their function is how they encode information. While many models have focused on neuronal firing rate as the primary code, this paper explores an alternative: the duration of neuronal responses. Specifically, the authors investigate whether the length of time a neuron is active (or inactive) in response to a stimulus can convey information about abstract features of that stimulus, particularly those related to reward probability. This is important because the BG are thought to act as "gates," modulating information flow to other brain areas, and the duration of this gating could be a critical parameter.

Methods

The researchers analyzed electrophysiological recordings from three key BG nuclei in monkeys performing a visuomotor task:

  • External Globus Pallidus (GPe): Involved in gating inputs to other BG structures.
  • Substantia Nigra pars Reticulata (SNr): A major output nucleus of the BG, projecting to the thalamus and superior colliculus.
  • Substantia Nigra pars Compacta (SNc): Dopaminergic neurons that modulate BG activity.

The monkeys were presented with visual cues that predicted different probabilities of receiving a reward. The core of the analysis involved:

  1. Identifying Responsive Neurons: Neurons showing a significant change in firing rate after cue presentation were identified.
  2. Defining Response Features: Beyond firing rate, they specifically analyzed:
  3. Response Duration: The length of time a neuron exhibited a sustained increase (positive response) or decrease (negative response) in firing rate. This was carefully defined for periods between 150-500 ms after cue onset.
  4. Extreme Amplitude: The maximal or minimal firing rate achieved during the response.
  5. Spike Count: The total number of spikes within a defined window.
  6. Information Theory Analysis: Mutual Information (MI) was used to quantify how much information each response feature (duration, amplitude, spike count) carried about the reward probability. MI is a powerful, non-parametric measure that can detect various forms of statistical dependence, going beyond simple correlations.
  7. Synergy and Redundancy Analysis: Measures like "normalized synergy" and "normalized contribution" were employed to understand how information from different response features combined. This helped determine if duration provided unique information or if it was redundant with other measures like spike count.

Datasets

The study utilized electrophysiological recordings from four monkeys performing a probabilistic visuomotor task. The dataset comprised:

  • 61 neurons from the GPe.
  • 56 neurons from the SNr.
  • 109 dopaminergic SNc neurons.

The data were collected during a specific cue epoch (first 500 ms after cue onset) to isolate the encoding of reward prospects from motor execution. No new benchmarks were introduced; the focus was on analyzing existing neural data with novel information-theoretic approaches.

Key Insights and Ideas

  • Duration as a Primary Code: The most striking finding is that response duration is not just a byproduct of firing rate but a significant carrier of information about abstract stimulus features (reward probability). This challenges the long-standing focus on firing rate alone.
  • Gating Mechanism Support: The results strongly support models of the basal ganglia that emphasize "gating" mechanisms. The duration of this "gate opening" (neuronal response) appears to be a key parameter for modulating information flow.
  • Differential Encoding in BG Nuclei: There are distinct encoding strategies across BG nuclei. SNc neurons primarily showed positive responses, while SNr and GPe neurons exhibited both increases and decreases in firing rate, with a notable proportion of duration-informative SNr and GPe neurons showing negative responses.
  • Negative Responses Carry Information: The paper highlights that decreased firing rates (negative responses) in SNr and GPe neurons are also informative about reward probability, and duration is particularly crucial for decoding this information.
  • Duration is More Informative Than Amplitude/Count (for some responses): For neurons showing negative responses, duration was found to be more informative than extreme amplitude or spike count. This suggests that the length of inhibition is a critical signal.
  • Synergy is Limited: Combining duration with other features like spike count or amplitude often resulted in redundancy rather than significant synergy, implying that duration captures a substantial portion of the available information.

Key Results

  • Widespread Duration Encoding: A significant percentage of responsive neurons in all three BG nuclei encoded reward probability through response duration: 42% in SNr, 26.9% in SNc, and 29.3% in GPe.
  • Negative Responses Dominate Duration Encoding in SNr/GPe: For duration-informative neurons, 68% of SNr neurons and 50% of GPe neurons encoded information via negative responses (decreased firing rate), compared to positive responses.
  • Duration Outperforms Other Features (in some cases): For negative responses, duration was more informative than extreme amplitude or spike count. For positive responses, duration was as informative as these other features.
  • Total Spike Count vs. Duration: While total spike count (over a fixed window) often carried as much or more information than duration, the authors argue that duration captures the essential information, and the total spike count's advantage might be due to its inclusion of non-responsive bins.
  • Faster Detection of Decreases: The appendix mathematically demonstrates that detecting a decrease in firing rate (negative response) is faster than detecting an increase of the same magnitude, potentially explaining why duration of inhibition is so informative.

Limitations and Future Work

  • Focus on Cue Epoch: The study was limited to the first 500 ms after cue onset, excluding later motor-related activity. This was a deliberate choice to isolate reward prospect encoding.
  • Positive SNc Responses: The SNc neurons studied primarily showed positive responses to cues predicting reward. The paper notes that other studies have explored duration encoding in SNc for omission of rewards, suggesting a broader role for duration encoding in different contexts.
  • Future Directions: The authors suggest that response duration might play a role in the competition between potential actions, influencing motor vigor through BG output. They also propose that the differential encoding of positive and negative responses by duration could be mechanistically linked to the gating function of the SNr, where closing the gate (positive response) might be less critical than opening it (negative response). Further research could explore how this duration-based code interacts with other BG computations and influences downstream targets.

Summary updated

Abstract

Several models have suggested that information transmission in the basal ganglia (BG) involves gating mechanisms, where neuronal activity modulates the extent of gate aperture and its duration. Here, we demonstrate that BG response duration is informative about a highly abstract stimulus feature and show that the duration of "gate opening" can indeed be used for information transmission through the BG. We analyzed recordings from three BG locations: the external part of the globus pallidus (GPe), the substantia nigra pars reticulata (SNr), and dopaminergic neurons from the substantia nigra pars compacta (SNc) during performance of a probabilistic visuomotor task. Most (>85%) of the neurons showed significant rate modulation following the appearance of cues predicting future reward. Trial-to-trial mutual information analysis revealed that response duration encoded reward prospects in many (42%) of the responsive SNr neurons, as well as in the SNc (26.9%), and the GPe (29.3%). Whereas the low-frequency discharge SNc neurons responded with only an increase in firing rate, SNr and GPe neurons with high-frequency tonic discharge responded with both increases and decreases. Conversely, many duration-informative neurons in SNr (68%) and GPe (50%) responded with a decreased rather than an increased rate. The response duration was more informative than the extreme (minimal or maximal) amplitude or spike count in responsive bins of duration-informative neurons. Thus response duration is not simply correlated with the discharge rate and can provide additional information to the target structures of the BG.
BibTeX
@article{parush2008encoding,
  author = {Naama Parush and David Arkadir and Alon Nevet and Genela Morris and Naftali Tishby and Israel Nelken and Hagai Bergman},
  title = {{Encoding by Response Duration in the Basal Ganglia}},
  journal = {Journal of Neurophysiology},
  year = {2008},
  doi = {10.1152/jn.90400.2008},
}
Figure