Final Report Summary - ACCDECMEM (Tracking accumulation processes in memory decisions)
Project objectives
The main goal of the project is to investigate how decisions based on remembered information unfold over time. In the grant application, I described the following objectives:
1. The first objective was to examine whether recognition memory and perceptual decision making tasks exhibit similar neural correlates of the evidence accumulation process
2. The second objective was to assess whether the neural correlate of inter-study- list-item similarity can be read out over the course of stimulus presentation during the encoding stage.
3. The third objective was developing a cognitive architecture that
captures the encoding, main- tenance and decision stages of
recognition memory performance. This model can explain pre- and
post-trial effects.
Work performed
To achieve these objectives, we have pursued several lines of work. First, we have developed a new task paradigm in which we can cleanly compare the neural correlates of decision making based on perceptual and based on memory information. We then collected EEG data in this task from a set of 30 healthy individuals, as well as from a set of 15 patients with electrodes implanted for seizure monitoring in epilepsy. We analyzed those data using methods that we have developed before to look for evidece of evidence accumulation. We did not find a strong signal that was directly ramping up (see results below), and therefore additionally developed a novel classifier analysis to track evidence.
In addition, we investigated the relationship between a cognitive
architecture (ACT-R) and brain oscillations. We then used this
architecture to create a model of distraction, which is the main
factor that affects decision making. This model will in the future be
used to describe how between-trial thinking affects decision making.
Results
We have demonstrated in an existing iEEG dataset of memory decisions that no clear correlates of evidence accumulation can be found. For that reason, we have developed a novel method in which we use a classifier to track how differences between low and high amounts of evidence develop over time. We presented preliminary results of this analysis at the Society for Mathematical Psychology in 2015, which demonstrated that some correlates could be found, in mostly frontal and temporal regions, and mostly different for perceptual versus memory decisions. We also have submitted a manuscript that shows that the centro-parietal potential (CPP) is sensitive to evidence accumulation in both the perceptual and the memory tasks.
As regards aim 3, we demonstrated that fronto-parietal 4–9 Hz theta oscillations covary with the activity of the working memory resource of the ACT-R cognitive architecture. in other words, fronto-parietal theta oscillations increase when activity of this resource increases, and fronto-parietal theta oscillations decrease when activity of this resource decreases. We are further investigating oscillatory synchronization in a simple perceptual decision making task (with randomly moving dots), in which we showed a predominantly fronto-posterior synchronization at the beginning of each trial, and a predominantly left-right synchronization at the end of each trial, around the response. This is one of the very first papers describing how the interactions between brain regions, manifesting as patterns of oscillatory synchronization, is involved in the decision making process.
We then went on to model explicitly what participants may be doing outside the decision interval: sometimes paying attention to the task, but sometimes mind-wandering. The model was implemented in the ACT-R cognitive architecture, and its main principle is that there is always a competition between two goals: task-performance and distraction. We showed that this model of distraction could fit data from a very simple Sustained-Attention-to-Response-Task that is often used to measure mind-wandering. We are now engaging in studies to further test the model in new mind-wandering paradigms, and in examining how this distraction can affect subsequent decision making.
Our work on evidence accumulation & memory mechanisms (aim 1) and cognitive architectures (aim 3) so far has resulted in several conference and journal publications. Moreover, the results have been presented in many talks at national and international universities and conferences.
Final Results & Impact
We expect that the current work will continue to bear fruit. As is clear from above, several manuscripts are still in preparation or under review, and those will be published in the coming years. In addition, the new direction we have taken, in creating computational models of distraction, will have important implications for society. Since distraction is an ever-growing problem in today’s hyperconnected world, modeling this could clarity what factors increase and decrease distractions, and how those affect decision making. We also plan to extend our work on relating the ACT-R cognitive architecture to brain oscillations in different tasks. For the latter project, we have recently obtained an internal grant from the University of Groningen, on which a PhD student has been appointed.
The main goal of the project is to investigate how decisions based on remembered information unfold over time. In the grant application, I described the following objectives:
1. The first objective was to examine whether recognition memory and perceptual decision making tasks exhibit similar neural correlates of the evidence accumulation process
2. The second objective was to assess whether the neural correlate of inter-study- list-item similarity can be read out over the course of stimulus presentation during the encoding stage.
3. The third objective was developing a cognitive architecture that
captures the encoding, main- tenance and decision stages of
recognition memory performance. This model can explain pre- and
post-trial effects.
Work performed
To achieve these objectives, we have pursued several lines of work. First, we have developed a new task paradigm in which we can cleanly compare the neural correlates of decision making based on perceptual and based on memory information. We then collected EEG data in this task from a set of 30 healthy individuals, as well as from a set of 15 patients with electrodes implanted for seizure monitoring in epilepsy. We analyzed those data using methods that we have developed before to look for evidece of evidence accumulation. We did not find a strong signal that was directly ramping up (see results below), and therefore additionally developed a novel classifier analysis to track evidence.
In addition, we investigated the relationship between a cognitive
architecture (ACT-R) and brain oscillations. We then used this
architecture to create a model of distraction, which is the main
factor that affects decision making. This model will in the future be
used to describe how between-trial thinking affects decision making.
Results
We have demonstrated in an existing iEEG dataset of memory decisions that no clear correlates of evidence accumulation can be found. For that reason, we have developed a novel method in which we use a classifier to track how differences between low and high amounts of evidence develop over time. We presented preliminary results of this analysis at the Society for Mathematical Psychology in 2015, which demonstrated that some correlates could be found, in mostly frontal and temporal regions, and mostly different for perceptual versus memory decisions. We also have submitted a manuscript that shows that the centro-parietal potential (CPP) is sensitive to evidence accumulation in both the perceptual and the memory tasks.
As regards aim 3, we demonstrated that fronto-parietal 4–9 Hz theta oscillations covary with the activity of the working memory resource of the ACT-R cognitive architecture. in other words, fronto-parietal theta oscillations increase when activity of this resource increases, and fronto-parietal theta oscillations decrease when activity of this resource decreases. We are further investigating oscillatory synchronization in a simple perceptual decision making task (with randomly moving dots), in which we showed a predominantly fronto-posterior synchronization at the beginning of each trial, and a predominantly left-right synchronization at the end of each trial, around the response. This is one of the very first papers describing how the interactions between brain regions, manifesting as patterns of oscillatory synchronization, is involved in the decision making process.
We then went on to model explicitly what participants may be doing outside the decision interval: sometimes paying attention to the task, but sometimes mind-wandering. The model was implemented in the ACT-R cognitive architecture, and its main principle is that there is always a competition between two goals: task-performance and distraction. We showed that this model of distraction could fit data from a very simple Sustained-Attention-to-Response-Task that is often used to measure mind-wandering. We are now engaging in studies to further test the model in new mind-wandering paradigms, and in examining how this distraction can affect subsequent decision making.
Our work on evidence accumulation & memory mechanisms (aim 1) and cognitive architectures (aim 3) so far has resulted in several conference and journal publications. Moreover, the results have been presented in many talks at national and international universities and conferences.
Final Results & Impact
We expect that the current work will continue to bear fruit. As is clear from above, several manuscripts are still in preparation or under review, and those will be published in the coming years. In addition, the new direction we have taken, in creating computational models of distraction, will have important implications for society. Since distraction is an ever-growing problem in today’s hyperconnected world, modeling this could clarity what factors increase and decrease distractions, and how those affect decision making. We also plan to extend our work on relating the ACT-R cognitive architecture to brain oscillations in different tasks. For the latter project, we have recently obtained an internal grant from the University of Groningen, on which a PhD student has been appointed.