Periodic Reporting for period 4 - Human Decisions (The Neural Determinants of Perceptual Decision Making in the Human Brain)
Période du rapport: 2019-11-01 au 2020-04-30
The term ‘decision making’ often calls to mind scenarios such as how to vote in an election or which course to take in college, yet even our perception of our sensory environment relies on a continuous stream of elementary judgments that can be equally life altering (e.g. is the traffic light red or green?). In the highly complex and dynamic environment that we inhabit, making accurate and timely decisions is a considerable challenge for the brain since the information it receives is almost always to some degree unreliable. Understanding how the brain overcomes this problem stands to illuminate principles of computation that extend to a wide range of cognitive operations (Shadlen & Kiani 2013), as well as holding the key to improving the diagnosis and treatment of the many brain disorders that impact on decision making abilities. This ERC project builds on the recent discovery of human brain signals that finely trace decision formation in real-time. These signals provide a direct view on the brain mechanisms that enable us to make our decisions and thus offers new opportunities to answer a number of major questions that were previously impossible to fully address. These include establishing the neural principles and processes that allow us to achieve the best balance between speed and accuracy in our decisions, to account for unreliable evidence and to represent our confidence in our decisions. An additional goal of this project is also to leverage these methodological and theoretical developments in order to gain deeper insights into the impact of natural cognitive aging on our decision making abilities.
In theme 1 of this project we explored how our decision making processes are adapted to different contextual demands. In a first project we examined how decisions are made when the timing and location of relevant sensory information is unpredictable. We showed that under such circumstances the brain relies on rapid visual responses evoked by target stimuli to trigger the evidence gathering process (Loughnane et al 2016, Curr Biol; Neweman et al 2017, J Neurosci; Devine et al 2018, eLife). Such processes had not previously been implicated in decision making and point to an important new avenue for future research. In a second project we examined how our decisions are adapted to meet time pressure demands. We showed that the brain employs a number of distinct adjustments including boosting the visual representation of relevant information, lowering the quantity of evidence needed to commit to a decision and an acceleration of the decision-reporting motor acts. These results, reported in Steinemann et al (2018, Nat Comms) reveal that the neural adjustments to time pressure are more diverse than the dominant models of decision making had suggested. In a further investigation we developed a new decision model which took account of these neural observations (Kelly et al, in press, Nat Hum Beh).
In theme 2 of this project we sought to establish how choice confidence is represented in the brain. We showed that the same neural evidence accumulation signals that contribute to our choices, continue to accumulate evidence after the choice has been made and this 'post-decisional' accumulation plays a critical role in facilitating error detection (Murphy et al, 2015, eLife). We capitalised on these discoveries to provide new, more detailed insights into the impact of natural aging on metacognition (Harty et al 2017, Neuroimage). Further studies examining relationships between laboratory based tests of choice confidence and real-world metrics of cognitive performance have been completed and a manuscript is being prepared.
In theme 3 of this project we examined in detail how natural aging impacts on decision making processes. A review of the literature by our group established that this has been a relatively understudies area and, in particular, neurophysiological analyses were lacking (Dully et al, 2018, Beh Brain Res). We ran a multi-experiment study on a group of older and younger adults who performed a set of perceptual decision making tasks. We fit a conventional model of decision making to their behavioural data and found effects that accorded with those commonly reported in the literature - most prominently that elderly individuals respond more slowly on cognitive tasks due to implementing a more conservative decision making strategy. However, accompanying neurophysiological analyses produced conflicting results, finding no differences in the quantity of evidence old versus young adults required to commit to a decision. We therefore developed a novel model that took into account our neural observations and it yielded important new insights. First, it indicated that older adults on average tended to accumulate evidence less efficiently on some, but not all, perceptual tasks. Most interestingly, the model also highlighted that older adults were better at maintaining attention throughout the task. Thus the model pointed to both positive and negative aging effects on perceptual decision making. The results were reported in McGovern et al (2018, Nat Hum Beh).
In theme 2 of this project we sought to establish how choice confidence is represented in the brain. We showed that the same neural evidence accumulation signals that contribute to our choices, continue to accumulate evidence after the choice has been made and this 'post-decisional' accumulation plays a critical role in facilitating error detection (Murphy et al, 2015, eLife). We capitalised on these discoveries to provide new, more detailed insights into the impact of natural aging on metacognition (Harty et al 2017, Neuroimage). Further studies examining relationships between laboratory based tests of choice confidence and real-world metrics of cognitive performance have been completed and a manuscript is being prepared.
In theme 3 of this project we examined in detail how natural aging impacts on decision making processes. A review of the literature by our group established that this has been a relatively understudies area and, in particular, neurophysiological analyses were lacking (Dully et al, 2018, Beh Brain Res). We ran a multi-experiment study on a group of older and younger adults who performed a set of perceptual decision making tasks. We fit a conventional model of decision making to their behavioural data and found effects that accorded with those commonly reported in the literature - most prominently that elderly individuals respond more slowly on cognitive tasks due to implementing a more conservative decision making strategy. However, accompanying neurophysiological analyses produced conflicting results, finding no differences in the quantity of evidence old versus young adults required to commit to a decision. We therefore developed a novel model that took into account our neural observations and it yielded important new insights. First, it indicated that older adults on average tended to accumulate evidence less efficiently on some, but not all, perceptual tasks. Most interestingly, the model also highlighted that older adults were better at maintaining attention throughout the task. Thus the model pointed to both positive and negative aging effects on perceptual decision making. The results were reported in McGovern et al (2018, Nat Hum Beh).
The above research achievements have offered a number of fundamental new insights into the neural mechanisms governing decision making while also developing new methodologies for finely tracing information processing at distinct levels along the sensorimotor hierarchy.
An overarching goal of the action has been to use neurophysiological human brain recordings to empirically test and refine the predictions of computational models. Until recently, work of this kind has been the sole preserve of invasive brain recording studies but owing to methodological advances achieved by the PI we do now have the ability to isolate non-invasive brain signals that can be directly related to specific neural computations underpinning decision formation. This paves the way for a powerful new neurally-informed modelling approach in humans, in which neural signal observations can mediate between alternative model variants otherwise difficult to discern through behavioural fits alone, and where necessary, guide the construction of new models that capture key realities of neural implementation that are overlooked in existing models . We have pioneered this approach in two recent papers that have yielded new insights into the impact of natural aging (McGovern et al 2018, Nat Hum Beh) and prior knowledge (Kelly et al in press, Nat Hum Beh) on decision making.
In so doing the work has also uncovered new mechanistic accounts of several long-studied human brain signals )including the P3b, Pe, CNV and the N2pc). At the same time, by adopting a mathematical and theoretical framework that has been well establishing in non-human neurophysiology research, the work has served to bridge longstanding methodological and conceptual gaps between human and non-human research (O'Connell et al 2018, Trends Neurosci).
The isolation of non-invasive target selection signals highlighted a potentially powerful new diagnostic metric for clinical disorders impacting on visuospatial orienting. An important and unique feature of the target selection signals identified is that they can be measured independently over each hemisphere, a capability which has potentially significant benefits for research on clinical brain disorders associated with selective attention impairments (e.g. unilateral neglect following stroke).
An overarching goal of the action has been to use neurophysiological human brain recordings to empirically test and refine the predictions of computational models. Until recently, work of this kind has been the sole preserve of invasive brain recording studies but owing to methodological advances achieved by the PI we do now have the ability to isolate non-invasive brain signals that can be directly related to specific neural computations underpinning decision formation. This paves the way for a powerful new neurally-informed modelling approach in humans, in which neural signal observations can mediate between alternative model variants otherwise difficult to discern through behavioural fits alone, and where necessary, guide the construction of new models that capture key realities of neural implementation that are overlooked in existing models . We have pioneered this approach in two recent papers that have yielded new insights into the impact of natural aging (McGovern et al 2018, Nat Hum Beh) and prior knowledge (Kelly et al in press, Nat Hum Beh) on decision making.
In so doing the work has also uncovered new mechanistic accounts of several long-studied human brain signals )including the P3b, Pe, CNV and the N2pc). At the same time, by adopting a mathematical and theoretical framework that has been well establishing in non-human neurophysiology research, the work has served to bridge longstanding methodological and conceptual gaps between human and non-human research (O'Connell et al 2018, Trends Neurosci).
The isolation of non-invasive target selection signals highlighted a potentially powerful new diagnostic metric for clinical disorders impacting on visuospatial orienting. An important and unique feature of the target selection signals identified is that they can be measured independently over each hemisphere, a capability which has potentially significant benefits for research on clinical brain disorders associated with selective attention impairments (e.g. unilateral neglect following stroke).