Phasic arousal suppresses biases in mice and humans across domains of decision-making

Jan Willem de Gee, Konstantinos Tsetsos, Lars Schwabe, Anne E Urai, David A. McCormick, Matthew J. McGinley, Tobias H. DonnerbioRxiv 2018 · DOI

TL;DR

This study reveals a fundamental principle in how our brains make decisions: transient bursts of arousal, measured by pupil dilation, actively suppress biases in how we accumulate evidence. This effect is remarkably consistent, observed across species (humans and mice), different sensory inputs (auditory), and even in memory-based decisions, suggesting a conserved mechanism for more rational decision-making.

What Problem Does This Paper Address?

We often make decisions by gradually accumulating evidence over time. However, our decisions aren't always purely rational; we can exhibit biases, like being overly cautious ("conservative") or too quick to commit ("liberal"). Previous research hinted that the brain's arousal systems, which fluctuate throughout the day, might play a role in modulating these biases, with rapid pupil dilations (a marker of phasic arousal) being linked to reduced biases in perceptual tasks. However, several key questions remained:

  • Universality of Bias Suppression: Does this arousal-induced bias suppression apply to both conservative and liberal biases, or just one type?
  • Cross-Species Generalization: Does this effect hold true in non-human animals, like mice, whose arousal systems are organized similarly to ours?
  • Domain Generalization: Does this principle extend beyond simple perceptual tasks to more complex, memory-based decisions?

This paper aims to answer these questions by investigating the relationship between phasic arousal and decision biases across species, sensory modalities, and decision domains.

Methods

The researchers employed a multi-pronged approach:

  • Behavioral Tasks: They designed auditory tone-in-noise detection tasks for both humans and mice, requiring them to accumulate evidence to make a "go" or "no-go" decision. Humans also performed a forced-choice task with manipulated signal probabilities (to induce liberal and conservative biases) and a memory-based recognition task.
  • Pupillometry: They precisely measured pupil diameter changes, specifically focusing on the rapid dilations (phasic arousal) that occur during task engagement. This is a well-established non-invasive proxy for noradrenergic activity in the brainstem, a key arousal system.
  • Computational Modeling: They used the drift-diffusion model (DDM), a standard framework for analyzing decision-making. The DDM breaks down a decision into components like evidence accumulation rate (drift rate), bias in accumulation (drift bias), starting point of evidence accumulation, and decision time. By fitting the DDM to behavioral data and allowing parameters to vary with pupil response, they could pinpoint how arousal affects the decision process.
  • Statistical Analysis: Mixed-effects linear modeling was used to quantify the relationships between pupil responses and behavioral metrics, controlling for individual differences and other factors.

Datasets

The study utilized data from several sources:

  • Newly collected human data: For the auditory go/no-go task and a forced-choice auditory detection task with manipulated signal probabilities.
  • Newly collected mouse data: For the auditory go/no-go task.
  • Previously published human data: For a memory-based recognition task.
  • Previously published mouse data: Re-analyzed for cross-species comparisons.

No new benchmarks were introduced; rather, existing datasets were used to address the specific research questions.

Key Insights and Ideas

  • Arousal as a "Bias Neutralizer": The most striking finding is that phasic arousal doesn't just suppress one type of bias; it actively pushes decisions towards neutrality, regardless of whether the bias is conservative or liberal. This suggests arousal acts as a general mechanism to reduce systematic deviations from optimal decision-making.
  • Cross-Species Conservation: The observed effects in mice closely mirror those in humans, highlighting that the fundamental role of phasic arousal in shaping decision biases is likely conserved across mammalian species. This is crucial for translating findings between animal models and human cognition.
  • Domain Generality: The principle extends beyond simple sensory perception to memory-based decisions, indicating that arousal's influence on evidence accumulation is a broad cognitive mechanism, not limited to immediate sensory input.
  • Mechanism of Action: The DDM analysis revealed that arousal primarily affects the drift bias – the tendency to favor one option over another during evidence accumulation – rather than altering the speed of evidence encoding or the decision threshold itself.

Key Results

  • Bias Reduction: In both humans and mice, trials with larger task-evoked pupil responses showed significantly reduced conservative biases in auditory detection tasks.
  • Liberal Bias Suppression: In humans, when liberal biases were experimentally induced, higher arousal also led to a reduction in this liberal bias, pushing choices towards neutrality.
  • Memory-Based Decisions: Similar bias suppression was observed in a visual recognition memory task, where higher arousal reduced both liberal and conservative biases.
  • Drift Bias Modulation: Computational modeling showed that pupil-linked arousal specifically modulated the drift bias parameter in the DDM, explaining the observed behavioral changes. For example, in the rare-signal condition, increased arousal shifted drift bias towards a more neutral value.
  • No Effect on Other Parameters: Phasic arousal had minimal or inconsistent effects on other DDM parameters like boundary separation, drift rate, or non-decision time, emphasizing its specific role in modulating accumulation bias.

Limitations and Future Work

  • Causality: While the study shows strong correlations, it primarily relies on correlational evidence. Future work could use direct manipulation of arousal systems (e.g., optogenetics in mice, pharmacological interventions in humans) to establish stronger causal links.
  • Neurochemical Specificity: While pupil dilation is linked to noradrenergic activity, the precise neurochemical and neural circuits mediating these effects across different decision domains and species require further investigation.
  • Motor Preparation vs. Arousal: The authors addressed concerns that pupil responses might be driven by motor preparation, particularly in go/no-go tasks. They used specific methods to mitigate this, but acknowledge that further experiments could more definitively disentangle these effects.
  • Tonic vs. Phasic Arousal: The study focused on phasic arousal. Understanding how tonic (baseline) arousal interacts with or differs from phasic arousal in modulating decision-making remains an important area for future research.

Summary updated

Abstract

Decisions are often made by accumulating ambiguous evidence over time. The brain’s arousal systems are activated during such decisions. In previous work in humans, we showed that evoked responses of arousal centers during decisions are reported by rapid dilations of the pupil, and predict a suppression of biases in the accumulation of decision-relevant evidence (de Gee et al. 2017). Here, we show that this arousal-related suppression in decision bias acts on both conservative and liberal biases, and generalizes across species (humans / mice), sensory systems (visual / auditory), and domains of decision-making (perceptual / memory-based). In challenging sound-detection tasks, the impact of spontaneous or experimentally induced choice biases was reduced under high arousal. Similar bias suppression occurred when evidence was drawn from memory. All these behavioral effects were explained by reduced evidence accumulation biases. Our results pinpoint a general principle of the interplay between phasic arousal and decision-making.
BibTeX
@misc{gee2018phasic,
  author = {Jan Willem de Gee and Konstantinos Tsetsos and Lars Schwabe and Anne E Urai and David A. McCormick and Matthew J. McGinley and Tobias H. Donner},
  title = {{Phasic arousal suppresses biases in mice and humans across domains of decision-making}},
  year = {2018},
  doi = {10.1101/447656},
  howpublished = {bioRxiv},
}
Figure