Reward modulates the effect of visual cortical microstimulation on perceptual decisions.
Reward modulates the effect of visual cortical microstimulation on perceptual decisions.
复制标题
作者：
Cicmil N;Cumming BG;Parker AJ;Krug K
有效的知觉决策依赖于将感官信息与对不同选择可获得的奖励的知识相结合。然而，目前尚不清楚奖励信号在哪里与知觉决策路径的多个阶段相互作用，也不知道这可能通过什么机制发生。我们结合了对视觉区域V5/MT中特定功能神经元组的微刺激和表现条件奖赏操作，而猴子则执行视觉辨别任务。当可获得的奖赏较大时，微刺激在将知觉选择转向受刺激神经元的刺激偏好方面效果较差。心理物理控制实验表明，这一结果不能用微刺激试验中反应策略的选择性变化来解释。一个用于分析行为绩效的有限累积决策模型显示，如果期望报酬调节了知觉决策的感觉表征阶段，那么预期报酬与微刺激的交互作用就可以被解释，除了在整合阶段众所周知的影响之外。DOI：http://dx.doi.org/10.7554/eLife.07832.001识别物体在三维(3D)空间中的移动方式依赖于被称为V5/MT的大脑区域。组成V5/MT区的神经元组成的群组，每个群组都对特定的运动方向和特定的3D深度有“偏好”。如果一组神经元检测到其首选的运动方向和3D深度，它将变得高度活跃。在一个被称为整合的过程中，大脑可以评估哪些神经元群是活跃的。这些信息然后可以用来计算出物体在太空中的运动。正确识别3D运动是否预期会产生有益的结果，可能会影响整合过程。这使得大脑在有大笔奖励的情况下增加了成功的可能性。到目前为止，人们认为V5/MT区的活动发生在整合之前，不受获得奖励的可能性的影响。除了被移动的物体“自然”刺激外，V5/MT神经元还可以通过一种名为微刺激的技术被“人工”激活，这种技术使用一个微小的电极来电刺激神经元群。微刺激可以使视觉感知偏向于人工激活神经元的运动和3D深度“偏好”。如果V5/MT神经元确实从大脑的其他区域接收到关于潜在奖励的信息，我们预计奖励将以不同的方式影响自然和人工刺激的神经活动。另一方面，如果V5/MT神经元没有收到任何关于奖励的信息，那么它们的活动是自然的还是人工的都将无关紧要；它们产生的信号将是相同的。Cicmil等人。让两只猴子完成一项任务，在这项任务中，他们可以因为正确识别三维圆柱体的旋转方向而获得奖励，并在一些试验中对特定的V5/MT神经元组进行微刺激。当获得更大的奖励时，微刺激不太能偏向猴子对3D圆柱体旋转方向的选择。总体而言，Cicmil等人的S研究结果表明，在整合发生之前，V5/MT神经元能够整合有关奖励的信息。下一步将记录V5/MT区的活动，以确切地调查这一过程是如何发生的。DOI：http://dx.doi.org/10.7554/eLife.07832.002
Effective perceptual decisions rely upon combining sensory information with knowledge of the rewards available for different choices. However, it is not known where reward signals interact with the multiple stages of the perceptual decision-making pathway and by what mechanisms this may occur. We combined electrical microstimulation of functionally specific groups of neurons in visual area V5/MT with performance-contingent reward manipulation, while monkeys performed a visual discrimination task. Microstimulation was less effective in shifting perceptual choices towards the stimulus preferences of the stimulated neurons when available reward was larger. Psychophysical control experiments showed this result was not explained by a selective change in response strategy on microstimulated trials. A bounded accumulation decision model, applied to analyse behavioural performance, revealed that the interaction of expected reward with microstimulation can be explained if expected reward modulates a sensory representation stage of perceptual decision-making, in addition to the better-known effects at the integration stage. DOI: http://dx.doi.org/10.7554/eLife.07832.001 Identifying how an object is moving in three-dimensional (3D) space depends upon a brain region known as V5/MT. The neurons that make up area V5/MT form groups that each have a ‘preference’ for a particular direction of movement and a particular 3D depth. If a group of neurons detects its preferred direction of movement and 3D depth, it will become highly active. The brain can assess which groups of neurons are active, in a process known as integration. This information can then be used to work out the object's movement in space. The process of integration can be influenced by whether a rewarding outcome is expected to result from identifying the 3D movement correctly. This allows the brain to increase its likelihood of success in situations where a large reward is on offer. Until now, it was thought that the activity in area V5/MT, which takes place before integration, was not affected by the likelihood of receiving a reward. As well as being ‘naturally’ stimulated by moving objects, the V5/MT neurons can also be ‘artificially’ activated by a technique called microstimulation, which uses a tiny electrode to electrically stimulate groups of neurons. Microstimulation can bias visual perception towards the movement and 3D depth ‘preference’ of the artificially activated neurons. If the V5/MT neurons do receive information about potential rewards from other areas of the brain, we would expect rewards to affect naturally and artificially stimulated neural activity in different ways. On the other hand, if the V5/MT neurons do not receive any information about reward, then it will not matter whether their activity is natural or artificial; the signal that they produce will be the same. Cicmil et al. gave two monkeys a task in which they could receive rewards for correctly identifying a three-dimensional cylinder's direction of rotation, and applied microstimulation to specific groups of V5/MT neurons on some of the trials. When a larger reward was available, microstimulation was less able to bias the monkeys' choices about the rotation direction of the 3D cylinders. Overall, Cicmil et al.'s results suggest that the V5/MT neurons are able to incorporate information about reward, before integration occurs. The next step will be to record the activity of area V5/MT to investigate exactly how this happens. DOI: http://dx.doi.org/10.7554/eLife.07832.002