Human dorsal anterior cingulate neurons signal conflict by amplifying task-relevant information

R. Becket Ebitz, Elliot H. Smith, Guillermo Horga, Catherine A. Schevon, Mark Yates, Guy M. McKhann, Matthew Botvinick, Sameer A. Sheth, Benjamin Y. HaydenbioRxiv 2020 · DOI

TL;DR

This study reveals that the dorsal anterior cingulate cortex (dACC) doesn't just detect conflict, but actively amplifies task-relevant information to help resolve it. By analyzing single neuron activity in humans performing a conflict task, researchers found that conflict boosts the neural representation of correct responses, rather than simply signaling the presence of conflict itself.

What Problem Does This Paper Address?

The dorsal anterior cingulate cortex (dACC) is known to be involved in conflict processing, showing increased activity when we face conflicting demands. However, the precise neural computations underlying this role have been debated. Three main hypotheses exist:

  1. Explicit Hypothesis: dACC neurons directly signal the presence of conflict, either as dedicated units or through a specific population coding axis.
  2. Epiphenomenal Hypothesis: The observed conflict-related activity is a byproduct of neurons firing for different actions or response predictions, not a direct computation of conflict itself.
  3. Amplification Hypothesis: Conflict influences dACC neurons by amplifying task-relevant information and suppressing irrelevant information, aiding in conflict resolution.

This paper aims to disentangle these hypotheses by examining single-neuron activity in humans during a conflict task.

Methods

The researchers employed a sophisticated approach combining neurophysiology and computational analysis:

  • Intracranial Recordings: Single-unit activity was recorded from the dACC and dorsolateral prefrontal cortex (dlPFC) of human patients undergoing monitoring for epilepsy or movement disorders.
  • Multi-Source Interference (MSI) Task: Participants performed a task designed to independently manipulate two types of conflict: Simon (motor-spatial) and Eriksen (perceptual flanker) conflict.
  • Targeted Dimensionality Reduction: This advanced technique was used to identify and analyze the "coding dimensions" within neuronal populations. These dimensions represent linear combinations of neuronal firing rates that best predict specific task variables (like correct responses or conflict). This allowed the researchers to move beyond simple correlations and understand the underlying computations.
  • Model Comparison: Statistical models were used to assess the additive and interactive effects of different conflict types on neuronal activity.

Datasets

The study utilized neural recordings from two cohorts of human patients:

  • Cohort 1: 7 patients with medically refractory epilepsy undergoing intracranial monitoring.
  • Cohort 2: 9 patients (8 with movement disorders, 1 with epilepsy) undergoing deep brain stimulation surgery or intracranial seizure monitoring.

These datasets provided single-unit activity from the dACC and dlPFC.

Key Insights and Ideas

  • Conflict as Amplification, Not Just Detection: The most striking finding is that dACC neurons don't just signal "conflict is happening." Instead, conflict seems to boost the neural representation of the correct, task-relevant information, making it easier to choose the right action.
  • Specialized Role of dACC: While dlPFC also shows some conflict-related activity, it's weaker and more heterogeneous. This suggests dACC has a more specialized role in conflict processing, potentially acting earlier in a hierarchical processing stream.
  • Rejection of Epiphenomenal Hypothesis: The data strongly argue against the idea that conflict signals are merely byproducts of other neural processes. The observed patterns of activity are not fully explained by the co-activation of neurons tuned for different responses.
  • No Abstract Conflict Signal: Contrary to the explicit hypothesis, the study found no evidence for a unified "abstract conflict signal" that is independent of the type of conflict. Different conflict types affected distinct pools of neurons.

Key Results

  • Increased dACC Activity: Conflict, particularly Eriksen conflict, led to a significant increase in average firing rates in dACC neurons.
  • Enhanced Correct Response Encoding: Population-level analysis showed that conflict increased the accuracy with which neural activity could predict the correct response. This effect was even stronger in conflict trials than in no-conflict trials.
  • Divergent Roles of dACC and dlPFC: dACC neurons showed stronger and more consistent responses to conflict compared to dlPFC neurons, which responded weakly and heterogeneously.
  • Distinct Neuron Pools: Conflict types (Simon vs. Eriksen) tended to affect different sets of neurons, suggesting a lack of a single, abstract conflict-detecting mechanism.

Limitations and Future Work

  • Translational Uncertainty: While the study uses human data, the authors acknowledge that the precise mechanisms might differ in non-human primates.
  • Source of Control: The study doesn't definitively pinpoint where the "cognitive control" for conflict originates. They propose two possibilities: either a distal region detects conflict and controls dACC neurons, or there's no single central executive, but rather a distributed system.
  • Future Directions: The authors suggest that future work, particularly modeling, will be needed to disambiguate the two possibilities for the origin of cognitive control. They also highlight the need to understand how executive processes modulate sensorimotor transformations more broadly.

Summary updated

Abstract

SUMMARY Hemodynamic activity in dorsal anterior cingulate cortex (dACC) correlates with conflict, suggesting it contributes to conflict processing. This correlation could be explained by multiple neural processes that can be disambiguated by population firing rates patterns. We used targeted dimensionality reduction to characterize activity of populations of single dACC neurons as humans performed a task that manipulates two forms of conflict. Although conflict enhanced firing rates, this enhancement did not come from a discrete population of domain-general conflict-encoding neurons, nor from a distinct conflict-encoding response axis. Nor was it the epiphenomenal consequence of simultaneous coactivation of action plans. Instead, conflict amplified the task-relevant information encoded across the neuronal population. Effects of conflict were weaker and more heterogeneous in the dorsolateral prefrontal cortex (dlPFC), suggesting that dACC’s role in conflict processing may be somewhat specialized. Overall, these results support the theory that conflict biases competition between sensorimotor transformation processes occurring in dACC.
BibTeX
@misc{ebitz2020human,
  author = {R. Becket Ebitz and Elliot H. Smith and Guillermo Horga and Catherine A. Schevon and Mark Yates and Guy M. McKhann and Matthew Botvinick and Sameer A. Sheth and Benjamin Y. Hayden},
  title = {{Human dorsal anterior cingulate neurons signal conflict by amplifying task-relevant information}},
  year = {2020},
  doi = {10.1101/2020.03.14.991745},
  howpublished = {bioRxiv},
}
Figure