What does the mean mean? A simple test for neuroscience.

Area of Science:

• Neuroscience
• Computational Neuroscience

Background:

• Trial-averaged metrics, such as tuning curves, are widely used to characterize neuronal activity.
• The computational relevance of these averaged responses remains a key question in neuroscience.

Purpose of the Study:

• To develop and apply a test assessing the relevance of trial-averaged neuronal responses to actual neuronal processing.
• To investigate the assumptions of reliability and behavioral relevance underlying the use of averaged metrics.

Main Methods:

• A novel test was designed to evaluate two assumptions: response reliability and behavioral relevance of single-trial responses to average templates.
• The test was applied to two datasets: optogenetic stimulation in mouse somatosensory cortex and electrophysiological recordings during a contrast discrimination task.

Main Results:

• In a highly controlled experiment (Dataset 1), both reliability and behavioral relevance assumptions were largely met.
• In a less restricted task (Dataset 2), neither assumption held, indicating trial-averaged responses were less relevant.
• Simulations suggested response diversity, not just reliability, explained performance differences.

Conclusions:

• Trial-averaged neuronal metrics may not accurately represent neuronal computation in tasks with less restricted behavior, where dynamic coding might prevail.
• Researchers are encouraged to use this test to validate the computational relevance of trial-averaged neuronal metrics in their specific experimental contexts.