Correlations of Neural Activity with Behavioral Kinematics during Active Tactile Decision Making

Ioannis Delis, Jacek P. Dmochowski, Paul Sajda, Qi WangSetn 2018 · DOI

TL;DR

This paper explores how our brain makes decisions when we actively explore the world through touch. By simultaneously measuring brain activity (EEG) and finger movements, they found that specific patterns of neural activity correlate with how we encode sensory information and integrate it to make a choice, offering a direct link between our physical exploration and cognitive processes.

What Problem Does This Paper Address?

Making decisions in the real world often involves actively sensing our environment, not just passively receiving information. Think about how you explore an object with your fingers to understand its shape or texture. This "active sensing" is crucial for gathering the information needed to make accurate and timely judgments. While we know that perceptual decisions involve integrating sensory evidence, and that our actions influence this evidence, the neural mechanisms underlying this active sensing process and its link to decision formation remain largely elusive. This paper aims to bridge this gap by investigating how the brain processes tactile information when we actively explore a surface to make a perceptual decision.

Methods

The researchers employed a novel methodology called EEG2Beh(aviour), an extension of a previous technique (Stim2EEG), to link neural activity (EEG) with sensorimotor behavior (finger kinematics). Here's a breakdown:

  • Active Tactile Discrimination Task: Participants used a haptic device (Pantograph) to explore virtual textures with varying amplitudes. They had to decide which of two regions had a higher amplitude.
  • Simultaneous Data Recording: During the task, they recorded both EEG signals (brain activity) and finger kinematics (movement velocity).
  • EEG2Beh(aviour) Algorithm: This core method uses Canonical Correlation Analysis (CCA) to find pairs of "components" – one representing a spatial pattern of EEG activity and the other a temporal pattern of finger movement velocity – that are maximally correlated. This effectively identifies how specific brain activity patterns are coupled with specific movement patterns.
  • Source Localization: To understand where in the brain these EEG components originate, they used a source reconstruction analysis (Brainstorm software) to map the EEG signals to specific brain regions.
  • Hierarchical Drift Diffusion Model (HDDM): To link these brain-behavior couplings to actual decision-making performance (accuracy and response time), they used an HDDM. This model decomposes decision-making into internal processes like evidence accumulation (drift rate) and stimulus encoding (non-decision time). Crucially, they used the trial-by-trial EEG2Beh coupling strengths as regressors in the HDDM, allowing them to see how these couplings influence specific decision parameters.

Datasets

The study used data collected from 13 healthy participants performing the tactile texture discrimination task. No new benchmark datasets were introduced.

Key Insights and Ideas

  • Active Exploration is Key: The study strongly suggests that how we explore a stimulus with our bodies provides direct insight into how our brain is forming a decision. It's not just about the sensory input, but the dynamic interaction between our actions and neural processing.
  • Brain Regions Have Specific Roles: Different brain regions are involved in distinct aspects of active tactile decision-making. The right lateral occipital cortex (LOC) seems to be involved in the initial encoding of the tactile stimulus, while the middle frontal gyrus (MFG) plays a role in integrating that evidence towards a choice.
  • Sensorimotor Behavior as a Window: The EEG2Beh method is powerful because it uses sensorimotor behavior (kinematics) as a proxy to understand underlying neural processes. This is particularly useful when direct neural recordings from deep brain structures are not feasible.
  • The "Non-Decision" Time is Not Passive: The study highlights that the non-decision time (which includes stimulus encoding and response production) is not a fixed, passive period. The fidelity of tactile sensory encoding, linked to the LOC, directly influences this duration.

Key Results

  • Three Significant EEG-Kinematic Components: The EEG2Beh analysis identified three pairs of significant EEG and kinematic components that consistently showed coupling across subjects.
  • LOC Modulates Non-Decision Time: The coupling strength of the EEG component localized to the right lateral occipital complex (LOC) was negatively correlated with non-decision time. This means stronger coupling (better brain-behavior synchronization) led to shorter non-decision times, suggesting more efficient stimulus encoding.
  • MFG Predicts Drift Rate: The coupling strength of the EEG component localized to the right middle frontal gyrus (MFG) was positively correlated with the drift rate. This indicates that stronger coupling in this region led to a faster rate of evidence accumulation towards a decision.
  • SMA Coupling without DDM Link: A third component localized to the premotor cortex and supplementary motor area (SMA) showed significant EEG-kinematic coupling but did not correlate with any HDDM parameters, suggesting its role might be more in motor execution rather than directly influencing decision parameters.
  • Behavioral Improvement: Task performance (accuracy and response time) significantly improved with increasing stimulus difference, as expected.

Limitations and Future Work

  • Limited Sample Size: While 13 subjects is reasonable for this type of study, larger sample sizes could further strengthen the findings.
  • Source Localization Limitations: The study used a standardized head model due to the unavailability of individual MRIs, which can introduce some inaccuracies in source localization.
  • Future Directions: The authors suggest that this approach could be extended to other sensory modalities and more complex decision-making tasks. They also emphasize the potential of using this method to understand how neurological disorders might affect active sensing and decision-making processes. The identification of distinct roles for the LOC and MFG in sensory encoding and evidence integration, respectively, opens avenues for targeted interventions or rehabilitation strategies.

Summary updated

BibTeX
@inproceedings{delis2018correlations,
  author = {Ioannis Delis and Jacek P. Dmochowski and Paul Sajda and Qi Wang},
  title = {{Correlations of Neural Activity with Behavioral Kinematics during Active Tactile Decision Making}},
  booktitle = {Setn},
  year = {2018},
  doi = {10.1145/3200947.3201027},
}
Figure