The Silent Period of Evidence Integration in Fast Decision Making

Johannes Rüter, Henning Sprekeler, Wulfram Gerstner, Michael H. HerzogPLoS 2013 · DOI

Abstract

In a typical experiment on decision making, one out of two possible stimuli is displayed and observers decide which one was presented. Recently, Stanford and colleagues (2010) introduced a new variant of this classical one-stimulus presentation paradigm to investigate the speed of decision making. They found evidence for "perceptual decision making in less than 30 ms". Here, we extended this one-stimulus compelled-response paradigm to a two-stimulus compelled-response paradigm in which a vernier was followed immediately by a second vernier with opposite offset direction. The two verniers and their offsets fuse. Only one vernier is perceived. When observers are asked to indicate the offset direction of the fused vernier, the offset of the second vernier dominates perception. Even for long vernier durations, the second vernier dominates decisions indicating that decision making can take substantial time. In accordance with previous studies, we suggest that our results are best explained with a two-stage model of decision making where a leaky evidence integration stage precedes a race-to-threshold process.

Introduction: The Speed of Choice

How fast do we actually make decisions? Traditional experiments measure reaction time (RT), but RT is a "noisy" metric because it bundles together sensory processing, the actual decision, and motor execution. Recent studies, notably Stanford et al. (2010), used a "compelled-response" paradigm to isolate the decision phase, suggesting that monkeys could make perceptual decisions in less than 30ms. This supported a "one-stage race-to-threshold" model, where sensory evidence immediately and directly drives a race toward a choice.

The authors of this paper questioned whether this rapid "rise time" truly reflects the entire decision process or if there is a hidden, earlier stage of evidence integration that is "invisible" to standard tests.

Methods: The Compelled-Response Paradigm

The researchers employed a compelled-response paradigm with human observers. In this setup, a "GO" signal appears before the stimulus. The subject must initiate a response immediately, effectively "guessing" until the stimulus appears. The time between stimulus onset and the response is the Processing Time (PT).

Experimental setup and models of decision making. Panel B shows the sequence of events in the compelled-response paradigm. Panel E illustrates the resulting tachometric function.

The study used two conditions:

  1. Single-Vernier (Control): A single vernier stimulus (two vertical lines with a horizontal offset) is shown.
  2. Two-Vernier (Experimental): A "vernier fusion" stimulus where one vernier is immediately followed by a second vernier with the opposite offset.

By varying the duration of the first vernier and measuring which direction the observer chose, the researchers could determine if the first stimulus had already "triggered" a decision before the second one arrived.

Key Insights: The Two-Stage Model

The study's findings challenge the notion that perceptual decisions are made in a single, ultra-fast "race" to a threshold. Instead, the authors propose a two-stage model:

  • Stage 1: Evidence Integration. Evidence is collected in a "leaky" visual buffer. This stage is "silent" because it doesn't trigger a motor commitment yet.
  • Stage 2: Decision Commitment. The integrated evidence is fed into a "race-to-threshold" process.
Components of Reaction Time (RT) in the decision-making process, highlighting the stages of sensory delay, evidence integration, and motor delay.

The "silent period" of evidence integration lasts at least 80ms. Because the second stimulus "cancels" or overrides the first, it proves that the brain is still integrating information long after a one-stage model would have already crossed the finish line.

Results

  • Single-Vernier Performance: For a single vernier, the tachometric function showed a rapid transition to accuracy, with a rise time of 63 ± 17 ms, similar to previous monkey studies.
  • Second-Stimulus Dominance: In the two-vernier condition (80ms of Vernier 1 followed by Vernier 2), the second vernier dominated the response. If the decision were truly made in <63ms, the first vernier should have won the "race."
  • Cancellation Effect: When both verniers were presented for 20ms each, they canceled each other out (50% performance), indicating that the integration process treats the two stimuli as a single fused event rather than two separate decision points.

Conclusion

The total decision-making process in this paradigm clearly exceeds 80ms, significantly longer than previously estimated by one-stage models. While the results are based on "vernier fusion," they suggest that decision-making involves a complex interplay between a leaky visual buffer and a rapid commitment phase. Future research should focus on identifying the neural correlates of this "buffer" stage to see where in the visual cortex this integration occurs before being handed off to motor-planning areas.

Summary updated

Abstract

In a typical experiment on decision making, one out of two possible stimuli is displayed and observers decide which one was presented. Recently, Stanford and colleagues (2010) introduced a new variant of this classical one-stimulus presentation paradigm to investigate the speed of decision making. They found evidence for "perceptual decision making in less than 30 ms". Here, we extended this one-stimulus compelled-response paradigm to a two-stimulus compelled-response paradigm in which a vernier was followed immediately by a second vernier with opposite offset direction. The two verniers and their offsets fuse. Only one vernier is perceived. When observers are asked to indicate the offset direction of the fused vernier, the offset of the second vernier dominates perception. Even for long vernier durations, the second vernier dominates decisions indicating that decision making can take substantial time. In accordance with previous studies, we suggest that our results are best explained with a two-stage model of decision making where a leaky evidence integration stage precedes a race-to-threshold process.
BibTeX
@inproceedings{ruter2013silent,
  author = {Johannes Rüter and Henning Sprekeler and Wulfram Gerstner and Michael H. Herzog},
  title = {{The Silent Period of Evidence Integration in Fast Decision Making}},
  booktitle = {PLoS},
  year = {2013},
  doi = {10.1371/journal.pone.0046525},
}
Figure