Evidence for separate processes underlying movement and decision vigor in a reward-oriented task

This study demonstrates that movement and decision vigor are governed by separate, independent time-cost signals rather than a single global utility, with their apparent coordination arising only when both processes are influenced by a shared temporal history.

The Gist

The Big Question: Is the Brain One Boss or Two?

Imagine you are at a crowded buffet. You have to decide two things:

1. How fast do you walk to the food station? (Movement)
2. How long do you stay at that station to pile your plate before moving to the next one? (Decision)

Scientists have long debated how the brain handles these two choices.

• Theory A (The "One Boss" Theory): The brain has a single "Manager" that looks at the whole picture. If walking is hard (too much effort), the Manager says, "Okay, let's walk slower, but we'll stay at the buffet longer to make up for it." Everything is linked.
• Theory B (The "Two Bosses" Theory): The brain has two separate managers. One handles walking, and one handles eating. They don't talk to each other much. If walking is hard, you just walk slower, but your eating habits stay exactly the same.

This paper set out to find out: Are we running on one global manager, or two separate ones?

---

The Experiment: The Robot Buffet

To test this, the researchers built a video game that acted like a digital buffet.

• The Setup: Participants sat in front of a screen with a robotic arm attached to their wrist.
• Phase 1: The Walk (Movement): They had to push against the robot arm to move a red dot across the screen to a green patch. Sometimes, the robot made it heavy (High Effort), and sometimes it was light (Low Effort).
• The Wait: Once they reached the patch, they had to wait a few seconds while it "filled up" with points.
• Phase 2: The Feast (Decision): Once the patch was full, they had to hold the robot arm steady against a "wind" pushing them away to collect points. The longer they stayed, the more points they got, but the points per second dropped over time (like a buffet where the best food is gone first). They could leave whenever they wanted.

The researchers changed the rules in different rounds:

1. Make the walk harder: Did they stay at the buffet longer to compensate?
2. Make the wait longer: Did they walk faster to make up for lost time?
3. Make the feast harder: Did they walk faster or slower?

The Results: The Managers Don't Talk

The results were surprising and pointed strongly to Theory B (Two Separate Bosses).

1. The "Heavy Walk" Test:
When the robot made the walk heavy (High Effort), people walked slower. But guess what? They did not stay at the buffet longer. They just accepted the slower walk and kept their eating habits exactly the same.

• Analogy: If your car engine is broken and you have to drive slower, you don't suddenly decide to stop at every gas station for an extra hour. You just drive slower.

2. The "Long Wait" Test:
When the researchers added a long delay before the feast started, people did stay at the buffet longer.

• Analogy: If you are forced to wait in a long line before you can eat, you feel like you've "wasted" time, so you decide to eat a bit longer once you finally get your food to make up for it.

3. The "Personality" Test:
The researchers looked at individual people.

• Some people were naturally fast walkers.
• Some people were naturally long stayers at the buffet.
• Crucially: The fast walkers were not necessarily the long stayers. A person who was super energetic about walking wasn't necessarily energetic about deciding when to leave. This proves the two skills are controlled by different parts of the brain.

The Secret Ingredient: The "Boredom Clock"

So, if the managers are separate, why do we sometimes see them working together? The paper found a hidden link: The Cost of Time (or the "Boredom Clock").

Imagine you have a clock in your head that starts ticking the moment a task begins.

• During the Wait: If there is a long delay (like waiting for the buffet to fill), that clock ticks loudly. You get "bored" or feel the "urgency" building up.
• The Effect: This "boredom" signal makes you want to move faster and stay longer to get your money's worth.
• The Catch: This signal is sensitive to delays you just experienced.
  • If you just waited in a long line, your "boredom clock" is high, so you stay at the buffet longer.
  • If you just walked slowly because the robot was heavy, your "boredom clock" didn't tick as much (because you were doing something), so you didn't change your eating habits.

The Conclusion

The brain doesn't use one giant calculator to balance effort and time for the whole day. Instead, it uses two separate optimization engines:

1. One engine decides how fast to move based on how hard it is to move.
2. Another engine decides how long to stay based on how much time was wasted waiting.

They only seem to be linked because they both listen to the same "Boredom Clock." If you are bored (because of a delay), both engines speed up. But if you are just tired (because of effort), only the movement engine slows down.

In short: We aren't one big machine trying to maximize efficiency globally. We are a collection of specialized parts that react to delays and effort independently, only occasionally syncing up when we feel the pressure of time running out.

Technical Summary

1. Problem Statement

The study addresses a fundamental debate in neuroscience and neuroeconomics regarding how the brain controls the "vigor" (speed and intensity) of movement and decision-making.

• The Common-Utility Hypothesis: Prominent theories (e.g., Shadmehr et al., 2016) suggest that movement and decision times are jointly regulated by a single control mechanism to maximize a global utility (the "capture rate": total reward minus effort, divided by total time). This implies a strong coupling: if movement effort increases, decision time should adjust to compensate to maintain the global rate.
• The Separate-Control Hypothesis: Alternative perspectives suggest that movement and decision vigor can be independently optimized.
• The Gap: Previous experimental results have been conflicting, often due to difficulties in dissociating the effects of time and effort across different behavioral phases. The authors aim to rigorously test these competing hypotheses by designing a task where time and effort can be manipulated independently in both movement and decision phases.

2. Methodology

Experimental Design
The researchers developed a foraging-like task using a robotic interface to create a controlled environment with isometric contractions.

• Participants: 44 healthy volunteers across a main experiment and two control experiments.
• Task Structure: Each trial consisted of two sequential phases:
  1. Reach Phase (Movement): Participants used their right hand to push against a robotic exoskeleton (isometric torque) to move a cursor from a start point to a target patch. Velocity was linearly proportional to applied torque ($v = k_r \mu$).
  2. Delay Phase: A fixed temporal delay occurred while the target patch "filled."
  3. Harvest Phase (Decision): Participants had to maintain a specific torque to counteract a virtual force field pushing the cursor out of the patch. Points accumulated over time but with diminishing returns (patch depletion). Participants decided when to stop harvesting and move to the next trial.
• Manipulations: The study utilized a block-wise design to independently manipulate:
  • Reach Effort: High vs. Low torque requirements to move the cursor.
  • Harvest Effort: High vs. Low torque requirements to hold the cursor in the patch.
  • Time Delays: High vs. Low delays introduced either between phases or before the reach phase.

Computational Modeling
Two distinct models were fitted to the behavioral data (median reach and harvest durations):

1. Common-Utility Model: Assumes a single global utility function ($J$) maximized over the trial. It integrates reward, effort (reach and harvest), and temporal discounting into one equation. It predicts that changes in one phase's cost should alter the optimal duration of the other phase.
2. Separate-Cost Model: Assumes two independent optimization problems.
   • Reach Cost ($C_r$): Balances effort cost against a "cost of time" (a sigmoidal function sensitive to preceding delays).
   • Harvest Cost ($C_h$): Balances effort cost, reward cost (opportunity cost of leaving early), and a separate "cost of time."
   • Key Feature: The cost of time in this model is sensitive to the offset (time elapsed since the start of the trial), allowing delays to indirectly influence the next phase without direct coupling.

Statistical Analysis

• Non-parametric tests (Wilcoxon signed-rank) for condition comparisons.
• Spearman correlations to assess inter-individual consistency (does a fast mover also make fast decisions?).
• Model comparison using the Akaike Information Criterion (AIC) to penalize model complexity.

3. Key Results

Behavioral Findings

• Decoupling of Effort:
  • Increasing Reach Effort significantly increased reach duration but had no significant effect on harvest duration.
  • Increasing Harvest Effort had no significant effect on either reach or harvest durations (participants maintained harvest duration despite higher effort, likely due to the high value of the reward).
• Effect of Delays:
  • Delays introduced before the reach phase slightly increased reach duration but significantly increased harvest duration.
  • Delays introduced between phases (after reach) significantly increased harvest duration but did not affect reach duration (as reach had already occurred).
• Inter-Individual Consistency:
  • There was strong consistency within phases: participants who were fast at reaching were consistently fast across conditions; those who harvested longer did so consistently.
  • Crucially: There was no significant correlation between an individual's reach duration and their harvest duration. Fast movers were not necessarily fast deciders, and vice versa.

Modeling Findings

• Superiority of the Separate-Cost Model: The Separate-Cost model provided a significantly better fit to the data (lower AIC and Residual Sum of Squares) compared to the Common-Utility model.
• Parameter Insights:
  • The Common-Utility model failed to predict the lack of cross-phase effects (e.g., it predicted harvest duration should change with reach effort, which it did not).
  • The Separate-Cost model successfully captured the data by treating the "cost of time" as a function of the time elapsed since the trial start.
  • The model revealed that the "cost of time" function is sigmoidal, with an inflection point that shifts based on preceding delays. This suggests that delays accumulate an exponential cost of time, which then transitions to a hyperbolic cost during the action itself, driving invigoration.

4. Key Contributions

1. Experimental Decoupling: The study successfully isolated movement and decision phases in a foraging task, demonstrating that they can be manipulated independently without the compensatory adjustments predicted by global utility maximization.
2. Evidence for Separate Control: The lack of correlation between individual movement vigor and decision vigor, combined with the failure of the common-utility model, provides strong evidence that the brain utilizes separate optimization processes for movement and decision-making.
3. Refined "Cost of Time" Mechanism: The authors propose that movement and decision vigor are linked only indirectly through a shared, time-sensitive signal (a "cost of time" or "urgency" signal) that accumulates during delays. This signal is sensitive to recent temporal history, explaining why delays affect subsequent vigor without requiring a global utility calculation.
4. Role of Reward vs. Effort: The findings suggest that in reward-oriented tasks, the value of the reward can override the influence of physical effort on decision duration, whereas movement duration remains sensitive to effort.

5. Significance

This paper challenges the prevailing "global capture rate" theory in neuroeconomics. It suggests that the brain does not necessarily compute a single, unified utility function for all volitional behaviors. Instead, it likely employs distinct control loops for movement and decision-making, which are only loosely coupled by a temporal signal reflecting the "boredom" or "urgency" of waiting.

This has implications for:

• Neuroscience: It suggests distinct neural substrates for movement vigor (potentially dorsal striatum) and decision vigor, which may only converge via temporal signals.
• Clinical Applications: Understanding these separate processes could help explain pathologies where movement and decision vigor are dissociated (e.g., in Parkinson's disease or depression, where apathy might affect one but not the other).
• Robotics and AI: It informs the design of autonomous agents, suggesting that separate optimization modules for action and decision, linked by a temporal cost signal, may be more robust than a single global utility maximizer in dynamic environments.