Time-adaptive modulation of evidence evaluation in rat posterior parietal cortex

Joint Public Review:

In this study, the authors sought to characterize the relationship between the timescales of evidence integration in an auditory change detection task and neural activity dynamics in the rat posterior parietal cortex (PPC), an area that has been implicated in the accumulation of sensory evidence. Using the state-of-the-art Neuropixel recording techniques, they identified two subpopulations of neurons whose firing rates were positively and negatively modulated by auditory clicks. The timescale of click-related response was similar to the behaviorally measured timescale for evidence evaluation. The click-related response of positively modulated neurons also depended on when the clicks were presented, which the authors hypothesized to reflect a time-dependent gain change to implement an urgency signal. Using muscimol injections to inactivate the PPC, they showed that PPC inactivation affected the rats' choices and reaction times.

There are several strengths of this study, including:

(1) Compelling evidence for short temporal integration in behavioral and neural data for this task.

(2) Well-executed and interpretable comparisons of psychophysical reverse correlation with single-trial, click-triggered neuronal analyses to relate behavior and neural activity.

(3) Inactivation experiments to test for causality.

(4) Characterization of neural subpopulations that allows for complex relationships between a brain region and behavior.

(5) Experimental evidence for an interesting way to use sensory gain change to implement urgency signals.

There are also some concerns, including:

(1) The work could be better contextualized. From a normative Bayesian perspective, the observed adaptation of timescales and gain aligns closely with optimal strategies for change detection in noisy streams: placing greater weight on recent sensory samples and lowering evidence requirements as decision urgency grows. However, the manuscript could go further in explicitly connecting the experimental findings to normative models, such as leaky accumulator or dynamic belief-updating frameworks. This would strengthen the broader impact of the work by making clear how the observed PPC dynamics instantiate computationally optimal strategies.

(2) It is unclear how the rats are performing the task, both in terms of the quality of performance (they only show hit rates, but the rats also seem to have high false alarm rates), and in terms of the underlying strategy that they seem to be using.

(3) A major conceptual weakness lies in the claim that PPC "dynamically modulates evidence evaluation in a time-adaptive manner to suit the behavioral demands of a free-response change detection task." To support this claim, it would require direct comparison of neural activity between two task demands, either in two tasks or in one task with manipulations that promote the adoption of different timescales.

(4) Some analyses of neural data are lacking or seem incomplete, without considering alternative interpretations.

(5) The muscimol inactivation results did not provide a clear interpretation about the link between PPC activity and decision performance.