< Back to previous page
Project
Dopaminergic reward signals selectively decrease fMRI activity in primate visual cortex
Looking at a specific object in a particular position in visual space reliably generates a pattern of neural activity in regions of the brain that process this visual information. We refer to this pattern of activity as the neural representation of that visual stimulus. In this thesiswe examined how a reward, which is a signal that does not originate in the retina (hence an extra-retinal signal), alters activity within the representation in the cortex of a visual stimulus when that stimulus is paired with a reward.
Evidence that stimulus-reward associations alterthe neural representation of a stimulus is suggested by behavioral experiments. These experiments demonstrated that a visual stimulus that predicts a reward becomes easier to detect and to discriminate. Importantly these behavioral improvements are specific to stimuli associated with reward and therefore suggest selective changes with the representation of reward-paired stimuli in visual cortex. Furthermore, clever experiments have shown that the improved processing (i.e. improved stimulus detection) of reward-paired stimuli can occur without attention. This finding demonstrates that the temporal coupling of stimulus and reward is sufficient to improve the processing of a stimulus.
These results lead one toask where reward information that is relayed to the visual processing regions of the brain comes from? One candidate is the dopaminergic systemof the ventral midbrain. These neurons respond more strongly when an event is more rewarding than expected, and they decrease their activity when an event is worse then expected. This response property is well described by prediction error, which is a measurement of how predictions change over time (i.e. the prediction can be larger or smaller than expected). This type of information is important to change behavior according todynamically changing environments or desires. In addition, increased activity in the ventral midbrain leads to dopamine release at a wide network of connection sites throughout the brain.
Importantly, when looking for reward related activity within brain regions that predominantly process visual information, one is faced with the difficulty of distinguishing changes in activity that are generated by rewards and those that are generated by attention toward rewarded stimuli. In an effort to isolate the effects of reward, we temporally separated the presentation of visual stimuli from the administration of a reward. We were inspired by studies demonstrating that rewards given at a time separate from stimulus-reward associations can actually strengthen these associations. Importantly, because these rewards strengthen stimulus preferences, we hypothesized that these rewards may selectively modulate activity within the representation of the stimulus.
To test this we designed an experiment that included trials which coupled a visual stimulus with a reward (stimulus-reward associations) and reward-only trials (without the presentation of any visual stimuli) while we monitored fMRI activity (a measure of brain activity) throughout the whole brain of monkeys. As hypothesized, wefound a selective change in activity during reward-only trials within the neural representation of the visual stimulus that had been coupled with reward. It is important to reiterate that the changes in activity we observed were during reward trials in which no visual stimulus was presented but nonetheless we found selective modulations in the visual regions of the brain.
We next aimed to determine what regulates the selective changes in activity we found in the visual cortex during the reward-only trials. We hypothesized that the prediction error-dependent signal of the ventral midbrain may be responsible. Thus we tested whether the magnitude and the location of reward modulations in the visual regions of the brain were dependent on prediction error. We did this by manipulating whether rewards were cued or not, altering the magnitude of rewards, varying the reward probability associated with cues and altering the predictability of rewards. In all of these different experiments, predictionerror described well the magnitude and location of the observed changesin activity during the reward-only trials. In one especially illustrative experiment, we found that the higher the probability of reward associated with a stimulus the stronger the changes in activity within the neural representation of that stimulus. Based on these results, we hypothesized that the influence of the prediction error on visual cortical activity during the reward-only trials depends upon dopaminergic signaling. To test this, we examined whether systemic application of a dopamine antagonist altered the reward-related activity in visual cortex. We found that the strength of the activity modulations was reduced, indicating dopamine was at least partially responsible for regulating the observed effect. Lastly we tested whether the reward-only trials affected the strength of the associations formed during the stimulus-reward association trials. As expected, we found a significant increase in the stimulus preference of the monkeys when stimulus- reward association trials were temporally surrounded by reward-only trials. Therefore these results show that in addition to modulating activity within visual cortex, the reward-onlytrials actually increase stimulus preference. Taken together, these results demonstrate that stimulus and reward information is brought together within the visual areas of the brain over relatively long time scales.More generally though, the findings of this thesis show that reward canexert very specific effects on visual cortex. The specificity of these signals may be used to selectively enhance the neural representation of rewarded stimuli.
Evidence that stimulus-reward associations alterthe neural representation of a stimulus is suggested by behavioral experiments. These experiments demonstrated that a visual stimulus that predicts a reward becomes easier to detect and to discriminate. Importantly these behavioral improvements are specific to stimuli associated with reward and therefore suggest selective changes with the representation of reward-paired stimuli in visual cortex. Furthermore, clever experiments have shown that the improved processing (i.e. improved stimulus detection) of reward-paired stimuli can occur without attention. This finding demonstrates that the temporal coupling of stimulus and reward is sufficient to improve the processing of a stimulus.
These results lead one toask where reward information that is relayed to the visual processing regions of the brain comes from? One candidate is the dopaminergic systemof the ventral midbrain. These neurons respond more strongly when an event is more rewarding than expected, and they decrease their activity when an event is worse then expected. This response property is well described by prediction error, which is a measurement of how predictions change over time (i.e. the prediction can be larger or smaller than expected). This type of information is important to change behavior according todynamically changing environments or desires. In addition, increased activity in the ventral midbrain leads to dopamine release at a wide network of connection sites throughout the brain.
Importantly, when looking for reward related activity within brain regions that predominantly process visual information, one is faced with the difficulty of distinguishing changes in activity that are generated by rewards and those that are generated by attention toward rewarded stimuli. In an effort to isolate the effects of reward, we temporally separated the presentation of visual stimuli from the administration of a reward. We were inspired by studies demonstrating that rewards given at a time separate from stimulus-reward associations can actually strengthen these associations. Importantly, because these rewards strengthen stimulus preferences, we hypothesized that these rewards may selectively modulate activity within the representation of the stimulus.
To test this we designed an experiment that included trials which coupled a visual stimulus with a reward (stimulus-reward associations) and reward-only trials (without the presentation of any visual stimuli) while we monitored fMRI activity (a measure of brain activity) throughout the whole brain of monkeys. As hypothesized, wefound a selective change in activity during reward-only trials within the neural representation of the visual stimulus that had been coupled with reward. It is important to reiterate that the changes in activity we observed were during reward trials in which no visual stimulus was presented but nonetheless we found selective modulations in the visual regions of the brain.
We next aimed to determine what regulates the selective changes in activity we found in the visual cortex during the reward-only trials. We hypothesized that the prediction error-dependent signal of the ventral midbrain may be responsible. Thus we tested whether the magnitude and the location of reward modulations in the visual regions of the brain were dependent on prediction error. We did this by manipulating whether rewards were cued or not, altering the magnitude of rewards, varying the reward probability associated with cues and altering the predictability of rewards. In all of these different experiments, predictionerror described well the magnitude and location of the observed changesin activity during the reward-only trials. In one especially illustrative experiment, we found that the higher the probability of reward associated with a stimulus the stronger the changes in activity within the neural representation of that stimulus. Based on these results, we hypothesized that the influence of the prediction error on visual cortical activity during the reward-only trials depends upon dopaminergic signaling. To test this, we examined whether systemic application of a dopamine antagonist altered the reward-related activity in visual cortex. We found that the strength of the activity modulations was reduced, indicating dopamine was at least partially responsible for regulating the observed effect. Lastly we tested whether the reward-only trials affected the strength of the associations formed during the stimulus-reward association trials. As expected, we found a significant increase in the stimulus preference of the monkeys when stimulus- reward association trials were temporally surrounded by reward-only trials. Therefore these results show that in addition to modulating activity within visual cortex, the reward-onlytrials actually increase stimulus preference. Taken together, these results demonstrate that stimulus and reward information is brought together within the visual areas of the brain over relatively long time scales.More generally though, the findings of this thesis show that reward canexert very specific effects on visual cortex. The specificity of these signals may be used to selectively enhance the neural representation of rewarded stimuli.
Date:12 Nov 2009 → 5 Apr 2013
Keywords:Visuality-driven, activity, Monkey
Disciplines:Neurosciences, Biological and physiological psychology, Cognitive science and intelligent systems, Developmental psychology and ageing
Project type:PhD project