Potential confounds in estimating trial-to-trial correlations between neuronal response and behavior using choice...

Area of Science:

• Neuroscience
• Computational Neuroscience
• Sensory Neuroscience

Background:

• Trial-to-trial fluctuations in sensory neuron responses correlate with perceptual reports, quantified by choice probability (CP).
• CP is a key tool for assessing neural contributions to behavior but often yields weak correlations requiring extensive data.
• Reliable CP estimation necessitates careful data analysis and rigorous collection controls.

Purpose of the Study:

• Identify potential data analysis confounds leading to biased choice probability (CP) estimates.
• Propose methods to mitigate bias in CP calculations.
• Highlight the impact of data analysis choices on interpreting neural-behavior correlations.

Main Methods:

• Analyzed potential biases in CP estimation arising from data normalization techniques.
• Investigated the influence of varying trial number ratios across conditions on CP estimates.
• Examined the effects of variable time intervals in quantifying neuronal responses on CP measurements.
• Simulated artifacts in reaction time tasks with time-varying neuronal responses and behavioral performance.

Main Results:

• Z-score normalization across stimulus conditions can underestimate CP when behavioral response category trial counts differ.
• Variable time intervals for response quantification can introduce artifacts, especially in reaction time tasks.
• Artifacts are pronounced when mean neuronal response and behavioral performance co-vary over time within trials.

Conclusions:

• Standard data analysis practices, like z-score normalization, can systematically bias CP estimates.
• Careful consideration of trial counts and measurement intervals is crucial for accurate neural-behavior correlation analysis.
• The findings underscore the need for rigorous data analysis protocols in neurophysiological studies to avoid misinterpreting neural contributions to behavior.