【Day35 文献精读】Causal contribution of primate auditory cortex to auditory perceptual decision-making

阅读文献：

Tsunada, J., Liu, A., Gold, J. et al. Erratum: Causal contribution of primate auditory cortex to auditory perceptual decision-making. Nat Neurosci 19, 642 (2016).

文献链接：

Erratum: Causal contribution of primate auditory cortex to auditory perceptual decision-making | Nature Neuroscience

本文为参加北京脑科学与类脑研究中心CIBR2022年冬季培训班Journal Club的文献阅读报告，感谢CIBR提供的学习机会和平台。

Contents

Abstract
Introduction
1. Research aim
2. Methods
3. Results
3.1 Behavioral performance
3.2 Recording-site localization
3.3 Neuronal stimulus sensitivity
3.4 Neuronal choice sensitivity
3.5 Microstimulation
4. Discussion
5. Conclusions

Abstract

1) Both ML and AL neural activity was modulated by the frequency content of the stimulus

2) The responses of AL(anterolateral) rather than ML(middle-lateral) belt regions of the auditory cortex neurons were systematically modulated by the monkeys’ choices and microstimulation.

3) AL directly and causally contributes sensory evidence to form this auditory decision.

Introduction

1. Research aim

Identify whether ML or AL auditory-driven responses are used as evidence to form decisions about the frequency content of auditory stimuli.

2. Methods

Recorded and manipulated ML and AL spiking activity in monkeys while made a difficult decision about whether a noisy stimulus contained more low- or high-frequency tone bursts (Fig. 1).

3. Results

3.1 Behavioral performance

	Psychometric data fit by logistic function	Chronometric data fit by DDM function	Spearman’s correlation coefficient
Dependent variable	Percent of high-frequency choices	Response time	/
Slope/ sensitivity	monkey T: 0.7 [0.5–0.7] monkey A: 0.8 [0.7-1.0]	monkey T: 0.6[0.5–0.7] monkey A: 0.7 [0.6-0.8]	R=0.70, P = 2.2 ×10^(-14)
Choice biases	monkey T: 13 [-5–31]% coherence monkey A: -22 [-30–7]% coherence	monkey T: 14 [-0.7–27]% coherence monkey A: -7 [-15–4]% coherence	R=0.96, P = 2.7 × 10^(-56)

1) From the psychometric functions (choices)

① High accuracy for high- and low-coherence stimuli

→ The monkeys were attentive and followed the rules of the task.

② Steepness of the logistic psychometric (choice) function:

monkey T: 0.7 [0.5–0.7] (median [interquartile range] across sessions)

monkey A: 0.8 [0.7-1.0]

→ The monkeys used relevant information from the auditory stimuli to inform their decisions.

③ Choice biases measured as the coherence value corresponding to 50% high-frequency choices from the logistic fits

monkey T: 13 [-5–31]% coherence

monkey A: -22 [-30–7]% coherence

→ The monkeys were also relatively unbiased, making roughly equal numbers of low- and high-frequency choices.

2) From the psychometric (choice) and chronometric (RT) functions

① Fit the two functions to drift-diffusion models (DDM) of decision-making with solid red curves.

② Non-decision time (NDT) was fit as a separate free parameter in the DDM for each of the two choices and included stimulus-encoding and motor-preparation times with dashed gray lines in chronometric functions.

③ Decision epoch (Decision time,DT) was defined by subtracting a NDT from RTs.

④ As absolute coherence increased, choices were both more accurate and faster.

⑤ Choice biases across sessions from the DDM fits

monkey T: 14 [-0.7–27]% coherence

monkey A: -7 [-15–4]% coherence

R(Spearman’s correlation coefficient between values from logistic and DDM fits)=0.96, P = 2.7 × 10^(-56)

⑥ Slopes of the DDM psychometric function:

monkey T: 0.6 [0.5–0.7]

monkey A: 0.7 [0.6-0.8]

R=0.70, P = 2.2 ×10^(-14)

→ these DDM fits implied relatively unbiased choices

3.2 Recording-site localization

1) From Fig. 3d

① ML frequency tuning increased at more posterior sites, whereas AL frequency tuning increased at more anterior sites.

② Neurons around the ML-AL border were tuned for low frequencies.

2) Our findings are robust to uncertainty about the specific location of the ML-AL border.

3.3 Neuronal stimulus sensitivity

1) From Fig. 4a: Coherence-dependent modulations at neuron level

Example ML neuron

Example AL neuron

Preferred frequency

high

low

Coherence-dependent modulations

yes

coh.↑ → firing rate↑

yes

coh.↓ → firing rate↑

slightly later response onsets and more sustained

→ Both ML and AL auditory-driven responses were modulated by the frequency content of the stimulus.

2) From Fig. 4b: Coherence-dependent modulations at population level

→ Both ML and AL showing sensitivity to signed coherence that was most prominent just after stimulus onset for ML (steeper curves), but was more persistent throughout stimulus presentation for AL.

3) From Fig.5a-b: Neurometric sensitivity to stimulus

① Receiver operating characteristic (ROC)-based ‘neurometric functions’

ROC value (%) : the probability that an ideal observer could use the spiking activity of an individual neuron to decide whether a given stimulus contained more high- or low-frequency tone bursts.

② Neurometric slopes (neurometric sensitivity): calculated from firing rates between stimulus onset and the inferred time of the decision commitment, median [IQR]

	ML	AL	Median difference
monkey T	0.3 [0.2–0.5]	0.3 [0.1–0.4]	P = 0.46, no significant difference
monkey A	0.4 [0.3–0.6]	0.4 [0.2–0.5]	P = 0.13, no significant difference

③ From Fig. 5b, the neurometric slopes tended to increase from just after stimulus onset until around the time of decision commitment, then turned to decresease.

④ From the insets of Fig. 5b, The neurometric slopes were similar for the two brain regions and the two monkeys (similar distribution of black bars).

→ ML and AL had similar neurometric sensitivity to the frequency content of the stimulus.

⑤ Median psychometric slopes (psychometric sensitivity) from all sessions:

both monkeys = 0.8 [0.7–1.0]

⑥ Neurometric slopes were slightly lower than the corresponding psychometric slopes for the two brain regions and the two monkeys (all P＜0.001).

→ On average, single-neuron ML and AL spiking activity was sensitive to stimulus coherence (neurometric sensitivity), but less so than psychometric sensitivity.

⑦ Either ML or AL activity could be pooled to improve sensitivity and provide the evidence needed to make the decision.

4) From Fig. 5c-e: session-by-session correlation between neurometric sensitivity and psychometric sensitivity

① the profiles were roughly similar in ML and AL

In both cases, they peaked around the inferred time of the decision commitment.

② the effects were statistically more reliable in AL

Especially in Fig. 5c, when aligned to stimulus onset, significant regression coefficients (red lines) appeared in AL but not in ML.

→There is a slightly closer association for AL versus ML activity and perceptual performance.

③ ML and AL stimulus-driven responses are similar, and either or both could, in principle, be used to inform the monkeys’ decisions.

3.4 Neuronal choice sensitivity

1) Choice probabilities of individual neurons

Quantifies the ability of an ROC-based ideal observer to use spiking activity to discriminate between low- and high-frequency.

2) From Fig. 6a: Neurons modulated by choices

ML neurons

AL neurons

Partial neurons

(black bars in Fig.6a)

20% (9 of 45)

P < 0.05, significant modulation

31% (17 of 55)

P < 0.05, significant modulation

Average level

P = 0.41

no significant modulation

P = 0.38

no significant modulation

① Certain ML and AL neurons were modulated by the monkeys’ choices.

② No systematic effect of choice on the spiking activity of populations of ML and AL neurons.

3) From Fig. 6b-e: Correlation between choice probability and neurometric sensitivity (slope) of individual neurons

① Significant regression coefficients appeared in AL but not in ML (red lines in Fig. 6b-d and Spearman's ρ≥0.50 with P＜0.01 in Fig. 6e)

→ The most sensitive AL neurons had task-driven activity that was positively related to the monkey’s choices.

② This positive correlation was evident during stimulus listening (Fig. 6b) around the inferred time of decision commitment (Fig. 6c) and persisted after the inferred time of movement initiation.

③ AL activity was more related to the monkeys’ decision-making behavior than ML activity.

3.5 Microstimulation

Preferred frequency	high-freq.	low-freq.
Expected direction	left	right
Joystick movement	contralateral	ipsilateral
Joystick movement	comparing to ML and AL (both from right hemisphere)

1) From Fig. 7a: Modulation of microstimulation at single-site level

Preferred frequency: high-freq. of monkey T

Single-site examples in AL

Single-site examples in ML

Shift

/choice

horizental translation of psychometric curve after stimulating

38% (P=0.001)

significant leftward shift

choices of high-freq.↑

2% (P＞0.05)

no significant shift

∆Slope

/sensitivity

-o.11 (P＞0.05)

no significant change of sensitivity

0.07 (P＞0.05)

no significant change of sensitivity

Preferred frequency: low-freq. of monkey A

Single-site examples in AL

Single-site examples in ML

Shift

/choice

horizental translation of psychometric curve after stimulating

13% (P=0.03)

significant rightward shift

choices of low-freq.↑

-7% (P＞0.05)

no significant shift

∆Slope

/sensitivity

-0.19 (P＞0.05)

no significant change of sensitivity

0.09 (P＞0.05)

no significant change of sensitivity

Four example sites illustrate:

① Microstimulation in AL (rather than ML) had a systematic effect on the monkeys’ choices, which means that the joystick movements tended to toward the expected side associated with the frequency after microstimulation.

→ Neurons with appropriate frequency tuning in AL, but not ML, provide sensory evidence used to form the decision.

② No significant changes were observed in the slope (sensitivity) of AL and ML after microstimulation.

2) From Fig. 7b: Modulation of microstimulation at population level

Consistent with the result ① at single-site level.

3) From Fig. 7c: Psychometric shift (Change in choice)

Microstimulation in AL shifted choices in the expected direction, median shift is 5% (P=0.01), but not in ML, 1% (P=0.45) across monkeys.

4) From Fig. 7d: Change in slope (sensitivity)

Both ML and AL microstimulation did not affect psychometric sensitivity (P＞0.05).

→ AL had a more direct and causal role than ML in the formation of the decisions

4. Discussion

1) AL and ML have similar stimulus-driven responses (results in 3.3 ), but different causal relationships with a perceptual decision (AL was more related to the monkeys’ decision-making behavior and microstimulation, results in 3.4 and 3.5).

2) We do not know how and where in the brain this decision is formed, but it is likely that downstream brain regions (for example, the ventrolateral prefrontal cortex) might accumulate
this evidence to form the decision.

3) Possible reasons for our findings inconsistent with previous researches that neural activity in auditory cortex is not reliably modulated by choice: differences in ① task designs, ② required decisions, ③ analysis of choice-related activity.

5. Conclusions

1) The ventral auditory pathway is functionally and causally involved in forming auditory perceptual decisions.

2) A simple, feedforward scheme might involve a representation of the acoustic features of a stimulus in the core auditory cortex and ML, which gets converted into task-relevant sensory evidence in AL.

3) The sensory evidence is used to form the decision in the ventrolateral prefrontal cortex.