The neural substrates of memory suppression: a FMRI exploration of directed forgetting

The directed forgetting paradigm is frequently used to determine the ability to voluntarily suppress information. However, little is known about brain areas associated with information to forget. The present study used functional magnetic resonance imaging to determine brain activity during the encoding and retrieval phases of an item-method directed forgetting recognition task with neutral verbal material in order to apprehend all processing stages that information to forget and to remember undergoes. We hypothesized that regions supporting few selective processes, namely recollection and familiarity memory processes, working memory, inhibitory and selection processes should be differentially activated during the processing of to-be-remembered and to-be-forgotten items. Successful encoding and retrieval of items to remember engaged the entorhinal cortex, the hippocampus, the anterior medial prefrontal cortex, the left inferior parietal cortex, the posterior cingulate cortex and the precuneus; this set of regions is well known to support deep and associative encoding and retrieval processes in episodic memory. For items to forget, encoding was associated with higher activation in the right middle frontal and posterior parietal cortex, regions known to intervene in attentional control. Items to forget but nevertheless correctly recognized at retrieval yielded activation in the dorsomedial thalamus, associated with familiarity-based memory processes and in the posterior intraparietal sulcus and the anterior cingulate cortex, involved in attentional processes.     

Dynamics of population code for working memory in the prefrontal cortex

Some neurons (delay cells) in the prefrontal cortex elevate their activities throughout the time period during which the animal is required to remember past events and prepare future behavior, suggesting that working memory is mediated by continuous neural activity. It is unknown, however, how working memory is represented within a population of prefrontal cortical neurons. We recorded from neuronal ensembles in the prefrontal cortex as rats learned a new delayed alternation task. Ensemble activities changed in parallel with behavioral learning so that they increasingly allowed correct decoding of previous and future goal choices. In well-trained rats, considerable decoding was possible based on only a few neurons and after removing continuously active delay cells. These results show that neural activity in the prefrontal cortex changes dynamically during new task learning so that working memory is robustly represented and that working memory can be mediated by sequential activation of different neural populations.     

When more means less: neural activity related to unsuccessful memory encoding

The neural correlates of memory encoding have been studied by contrasting neural activity elicited by items at the time of learning according to whether they were later remembered or forgotten [1]. Previous studies have focused on regions where neural activity is greater for subsequently remembered items [2-8]. Here, we describe regions where activity is greater for subsequently forgotten items. In two experiments that employed the same incidental learning task, activity in an overlapping set of cortical regions (posterior cingulate, inferior and medial parietal, and dorsolateral prefrontal) was associated with failure on a subsequent memory test.     

The avian 'prefrontal cortex' and cognition

Both mammals and birds can flexibly organize their behavior over time. In mammals, the mental operations generating this ability are called executive functions and are associated with the prefrontal cortex. The corresponding structure in birds is the nidopallium caudolaterale. Anatomical, neurochemical, electrophysiological and behavioral studies show these structures to be highly similar. The avian forebrain displays no lamination that corresponds to the mammalian neocortex, hence lamination does not seem to be a requirement for higher cognitive functions. Because all other aspects of the neural architecture of the mammalian and the avian prefrontal areas are extremely comparable, the freedom to create different neural architectures that generate prefrontal functions seems to be very limited.     

Mirroring intentional forgetting in a shared-goal learning situation

Background

Intentional forgetting refers to the surprising phenomenon that we can forget previously successfully encoded memories if we are instructed to do so. Here, we show that participants cannot only intentionally forget episodic memories but they can also mirror the “forgetting performance” of an observed model.

Methodology/Principal Findings

In four experiments a participant observed a model who took part in a memory experiment. In Experiment 1 and 2 observers saw a movie about the experiment, whereas in Experiment 3 and 4 the observers and the models took part together in a real laboratory experiment. The observed memory experiment was a directed forgetting experiment where the models learned two lists of items and were instructed either to forget or to remember the first list. In Experiment 1 and 3 observers were instructed to simply observe the experiment (“simple observation” instruction). In Experiment 2 and 4, observers received instructions aimed to induce the same learning goal for the observers and the models (“observation with goal-sharing” instruction). A directed forgetting effect (the reliably lower recall of to-be-forgotten items) emerged only when models received the “observation with goal-sharing” instruction (P<.001 in Experiment 2, and P<.05 in Experiment 4), and it was absent when observers received the “simple observation” instruction (P>.1 in Experiment 1 and 3).

Conclusion

If people observe another person with the same intention to learn, and see that this person is instructed to forget previously studied information, then they will produce the same intentional forgetting effect as the person they observed. This seems to be a an important aspect of human learning: if we can understand the goal of an observed person and this is in line with our behavioural goals then our learning performance will mirror the learning performance of the model.     

Recurrent neural networks of integrate-and-fire cells simulating short-term memory and wrist movement tasks derived from continuous dynamic networks

Dynamic recurrent neural networks composed of units with continuous activation functions provide a powerful tool for simulating a wide range of behaviors, since the requisite interconnections can be readily derived by gradient descent methods. However, it is not clear whether more realistic integrate-and-fire cells with comparable connection weights would perform the same functions. We therefore investigated methods to convert dynamic recurrent neural networks of continuous units into networks with integrate-and-fire cells. The transforms were tested on two recurrent networks derived by backpropagation. The first simulates a short-term memory task with units that mimic neural activity observed in cortex of monkeys performing instructed delay tasks. The network utilizes recurrent connections to generate sustained activity that codes the remembered value of a transient cue. The second network simulates patterns of neural activity observed in monkeys performing a step-tracking task with flexion/extension wrist movements. This more complicated network provides a working model of the interactions between multiple spinal and supraspinal centers controlling motoneurons. Our conversion algorithm replaced each continuous unit with multiple integrate-and-fire cells that interact through delayed "synaptic potentials". Successful transformation depends on obtaining an appropriate fit between the activation function of the continuous units and the input-output relation of the spiking cells. This fit can be achieved by adapting the parameters of the synaptic potentials to replicate the input-output behavior of a standard sigmoidal activation function (shown for the short-term memory network). Alternatively, a customized activation function can be derived from the input-output relation of the spiking cells for a chosen set of parameters (demonstrated for the wrist flexion/extension network). In both cases the resulting networks of spiking cells exhibited activity that replicated the activity of corresponding continuous units. This confirms that the network solutions obtained through backpropagation apply to spiking networks and provides a useful method for deriving recurrent spiking networks performing a wide range of functions.     

Directed Forgetting of Negative Self-Referential Information Is Difficult: An fMRI Study

A large body of evidence suggested that both emotion and self-referential processing can enhance memory. However, it remains unclear how these two factors influence directed forgetting. This study speculates that directed forgetting of negative self-referential memory is more difficult than forgetting of other-referential memory. To verify this speculation, we combined the directed forgetting paradigm with the self-reference task. The behavioral result suggested that although both self-referential and other-referential information can be directly forgotten, less self-referential information can be forgotten than other-referential information. At the neural level, the forget instruction strongly activated the frontal cortex, suggesting that directed forgetting is not memory decay but an active process. In addition, compared with the negative other-referential information, forgetting of the negative self-referential information were associated with a more widespread activation, including the orbital frontal gyrus (BA47), the inferior frontal gyrus (BA45, BA44), and the middle frontal gyrus. Our results suggest that forgetting of the self-referential information seems to be a more demanding and difficult process.     

Helpful amnesia: Neuroscience unravels the role of forgetting as a prerequisite for establishing new long‐term memories

As neuroscience has been analysing the mechanisms behind long‐term memory, it demonstrated that forgetting is crucial for being able to remember.

“To be able to forget means sanity,” explained American writer Jack London (The Star Rover) referring to our sometimes infuriating inability to recall past events. In fact, being able to remember everything ever said and done might drive even the strongest mind insane. It is through forgetting and letting go of memories that the brain is able to acquire fresh impressions and new experience to move on, instead of being mired in the past.The importance of forgetting also fits well into and has inspired research to understand the molecular and cognitive basis of long‐term memory and how all the components and processes fit together. This puzzle of what is being remembered and why has been a long‐standing challenge for neuroscience; while progress has been made identifying more of the mechanisms and some of the existential drivers of memory formation, it is only recently that work has begun analyzing how these interact in animal models, often focusing on how the brain “decides” which things to remember and which things to send into oblivion.

This puzzle of what is being remembered and why has been a long‐standing challenge for neuroscience…

As a result, the field now sees collaboration across disciplines, driven by the realization that different parts and processes all play a part in memory formation and long‐term consolidation. This has led to one tangible if still tentative conclusion about long‐term memory, namely that forgetting occurs through loss of retrieval capability rather than erasure. It would appear to confirm the observation attributed to German philosopher Friedrich Nietzsche that “the existence of forgetting has never been proved: we only know that some things do not come to our mind when we want them to.”

The existence of forgetting has never been proved: we only know that some things do not come to our mind when we want them to.

     

Neural basis for successful encoding and retrieval of prospective memory

Prospective memory (PM) refers to memory for future intentions. Difference due to memory (Dm effect) is the difference in neural activity related to stimuli that were subsequently remembered or forgotten. Using event-related potentials (ERPs), the present study investigated the Dm effect for PM using a subsequent task-switching paradigm. The results showed that a Dm effect of ERP P150 was more positive-going for later PM hit trials than for later PM forgotten trials during 100–200 ms. This Dm effect may reflect the process for the production of future intention or the process for attention. Consistent with previously reported Dm effects of other types of memory, we found that the fbN2 (250–280 ms) and late positivity component (400–700 ms) were stronger in later PM hit trials than in forgotten trials. The fbN2 was evoked by Chinese characters. The late positivity component was related to the precise encoding process. In conclusion, because of the early P150, PM encoding appears to be somewhat different from previously identified Dm effects. However, further research is needed. Our findings reveal that Dm effects of PM share similar characteristics with known Dm effects of other types of episodic memory after the very early stage of neural processing.     

Towards an executive without a homunculus: computational models of the prefrontal cortex/basal ganglia system

The prefrontal cortex (PFC) has long been thought to serve as an 'executive' that controls the selection of actions and cognitive functions more generally. However, the mechanistic basis of this executive function has not been clearly specified often amounting to a homunculus. This paper reviews recent attempts to deconstruct this homunculus by elucidating the precise computational and neural mechanisms underlying the executive functions of the PFC. The overall approach builds upon existing mechanistic models of the basal ganglia (BG) and frontal systems known to play a critical role in motor control and action selection, where the BG provide a 'Go' versus 'NoGo' modulation of frontal action representations. In our model, the BG modulate working memory representations in prefrontal areas to support more abstract executive functions. We have developed a computational model of this system that is capable of developing human-like performance on working memory and executive control tasks through trial-and-error learning. This learning is based on reinforcement learning mechanisms associated with the midbrain dopaminergic system and its activation via the BG and amygdala. Finally, we briefly describe various empirical tests of this framework.     

Slow manifolds within network dynamics encode working memory efficiently and robustly

Working memory is a cognitive function involving the storage and manipulation of latent information over brief intervals of time, thus making it crucial for context-dependent computation. Here, we use a top-down modeling approach to examine network-level mechanisms of working memory, an enigmatic issue and central topic of study in neuroscience. We optimize thousands of recurrent rate-based neural networks on a working memory task and then perform dynamical systems analysis on the ensuing optimized networks, wherein we find that four distinct dynamical mechanisms can emerge. In particular, we show the prevalence of a mechanism in which memories are encoded along slow stable manifolds in the network state space, leading to a phasic neuronal activation profile during memory periods. In contrast to mechanisms in which memories are directly encoded at stable attractors, these networks naturally forget stimuli over time. Despite this seeming functional disadvantage, they are more efficient in terms of how they leverage their attractor landscape and paradoxically, are considerably more robust to noise. Our results provide new hypotheses regarding how working memory function may be encoded within the dynamics of neural circuits.     

Neural correlates of knowledge: stable representation of stimulus associations across variations in behavioral performance

Behavioral responses to a sensory stimulus are often guided by associative memories. These associations remain intact even when other factors determine behavior. The substrates of associative memory should therefore be identifiable by neuronal responses that are independent of behavioral choices. We tested this hypothesis using a paired-associates task in which monkeys learned arbitrary associations between pairs of visual stimuli. We examined the activity of neurons in inferior temporal cortex as the animals prepared to choose a remembered stimulus from a visual display. The activity of some neurons (22%) depended on the monkey's behavioral choice; but for a novel class of neurons (54%), activity reflected the stimulus that the monkey was instructed to choose, regardless of the behavioral response. These neurons appear to represent memorized stimulus associations that are stable across variations in behavioral performance. In addition, many neurons (74%) were modulated by the spatial arrangement of the stimuli in the display.     

What,where and when: deconstructing memory

The ability of animals to remember the what, where and when of a unique past event is used as an animal equivalent to human episodic memory. We currently view episodic memory as reconstructive, with an event being remembered in the context in which it took place. Importantly, this means that the components of a what, where, when memory task should be dissociable (e.g. what would be remembered to a different degree than when). We tested this hypothesis by training hummingbirds to a memory task, where the location of a reward was specified according to colour (what), location (where), and order and time of day (when). Although hummingbirds remembered these three pieces of information together more often than expected, there was a hierarchy as to how they were remembered. When seemed to be the hardest to remember, while errors relating to what were more easily corrected. Furthermore, when appears to have been encoded as a combination of time of day and sequence information. As hummingbirds solved this task using reconstruction of different memory components (what, where and when), we suggest that similar deconstructive approaches may offer a useful way to compare episodic and episodic-like memories.     

Primary and Secondary Rewards Differentially Modulate Neural Activity Dynamics during Working Memory

Background

Cognitive control and working memory processes have been found to be influenced by changes in motivational state. Nevertheless, the impact of different motivational variables on behavior and brain activity remains unclear.

Methodology/Principal Findings

The current study examined the impact of incentive category by varying on a within-subjects basis whether performance during a working memory task was reinforced with either secondary (monetary) or primary (liquid) rewards. The temporal dynamics of motivation-cognition interactions were investigated by employing an experimental design that enabled isolation of sustained and transient effects. Performance was dramatically and equivalently enhanced in each incentive condition, whereas neural activity dynamics differed between incentive categories. The monetary reward condition was associated with a tonic activation increase in primarily right-lateralized cognitive control regions including anterior prefrontal cortex (PFC), dorsolateral PFC, and parietal cortex. In the liquid condition, the identical regions instead showed a shift in transient activation from a reactive control pattern (primary probe-based activation) during no-incentive trials to proactive control (primary cue-based activation) during rewarded trials. Additionally, liquid-specific tonic activation increases were found in subcortical regions (amygdala, dorsal striatum, nucleus accumbens), indicating an anatomical double dissociation in the locus of sustained activation.

Conclusions/Significance

These different activation patterns suggest that primary and secondary rewards may produce similar behavioral changes through distinct neural mechanisms of reinforcement. Further, our results provide new evidence for the flexibility of cognitive control, in terms of the temporal dynamics of activation.     

Effect of expected reward magnitude on the response of neurons in the dorsolateral prefrontal cortex of the macaque

The dorsolateral prefrontal cortex plays a critical role in guiding actions that ensue seconds after an instruction. We recorded from neurons in area 46 and the frontal eye field (FEF) while monkeys performed a memory-guided eye movement task. A visual cue signaled whether a small or large liquid reward would accompany a correct response. Many neurons in area 46 responded more when the monkey expected a larger reward. Reward-related enhancement was evident throughout the memory period and was most pronounced when the remembered target appeared in the neuron's response field. Enhancement was not present in the FEF. The mixture of neural signals representing spatial working memory and reward expectation appears to be a distinct feature of area 46.     

Real-Time State Estimation in a Flight Simulator Using fNIRS

Working memory is a key executive function for flying an aircraft. This function is particularly critical when pilots have to recall series of air traffic control instructions. However, working memory limitations may jeopardize flight safety. Since the functional near-infrared spectroscopy (fNIRS) method seems promising for assessing working memory load, our objective is to implement an on-line fNIRS-based inference system that integrates two complementary estimators. The first estimator is a real-time state estimation MACD-based algorithm dedicated to identifying the pilot’s instantaneous mental state (not-on-task vs. on-task). It does not require a calibration process to perform its estimation. The second estimator is an on-line SVM-based classifier that is able to discriminate task difficulty (low working memory load vs. high working memory load). These two estimators were tested with 19 pilots who were placed in a realistic flight simulator and were asked to recall air traffic control instructions. We found that the estimated pilot’s mental state matched significantly better than chance with the pilot’s real state (62% global accuracy, 58% specificity, and 72% sensitivity). The second estimator, dedicated to assessing single trial working memory loads, led to 80% classification accuracy, 72% specificity, and 89% sensitivity. These two estimators establish reusable blocks for further fNIRS-based passive brain computer interface development.     

Time Course of Functional Connectivity in Primate Dorsolateral Prefrontal and Posterior Parietal Cortex during Working Memory

The dorsolateral prefrontal and posterior parietal cortex play critical roles in mediating attention, working memory, and executive function. Despite proposed dynamic modulation of connectivity strength within each area according to task demands, scant empirical data exist about the time course of the strength of effective connectivity, particularly in tasks requiring information to be sustained in working memory. We investigated this question by performing time-resolved cross-correlation analysis for pairs of neurons recorded simultaneously at distances of 0.2–1.5 mm apart of each other while monkeys were engaged in working memory tasks. The strength of effective connectivity determined in this manner was higher throughout the trial in the posterior parietal cortex than the dorsolateral prefrontal cortex. Significantly higher levels of parietal effective connectivity were observed specifically during the delay period of the task. These differences could not be accounted for by differences in firing rate, or electrode distance in the samples recorded in the posterior parietal and prefrontal cortex. Differences were present when we restricted our analysis to only neurons with significant delay period activity and overlapping receptive fields. Our results indicate that dynamic changes in connectivity strength are present but area-specific intrinsic organization is the predominant factor that determines the strength of connections between neurons in each of the two areas.     

Clinical and empirical applications of the Rey-Osterrieth Complex Figure Test

The Rey-Osterrieth Complex Figure Test (ROCF), which was developed by Rey in 1941 and standardized by Osterrieth in 1944, is a widely used neuropsychological test for the evaluation of visuospatial constructional ability and visual memory. Recently, the ROCF has been a useful tool for measuring executive function that is mediated by the prefrontal lobe. The ROCF consists of three test conditions: Copy, Immediate Recall and Delayed Recall. At the first step, subjects are given the ROCF stimulus card, and then asked to draw the same figure. Subsequently, they are instructed to draw what they remembered. Then, after a delay of 30 min, they are required to draw the same figure once again. The anticipated results vary according to the scoring system used, but commonly include scores related to location, accuracy and organization. Each condition of the ROCF takes 10 min to complete and the overall time of completion is about 30 min.     

Lmo Mutants Reveal a Novel Role for Circadian Pacemaker Neurons in Cocaine-Induced Behaviors

Drosophila has been developed recently as a model system to investigate the molecular and neural mechanisms underlying responses to drugs of abuse. Genetic screens for mutants with altered drug-induced behaviors thus provide an unbiased approach to define novel molecules involved in the process. We identified mutations in the Drosophila LIM-only (LMO) gene, encoding a regulator of LIM-homeodomain proteins, in a genetic screen for mutants with altered cocaine sensitivity. Reduced Lmo function increases behavioral responses to cocaine, while Lmo overexpression causes the opposite effect, reduced cocaine responsiveness. Expression of Lmo in the principal Drosophila circadian pacemaker cells, the PDF-expressing ventral lateral neurons (LN_vs), is sufficient to confer normal cocaine sensitivity. Consistent with a role for Lmo in LN_v function, Lmo mutants also show defects in circadian rhythms of behavior. However, the role for LN_vs in modulating cocaine responses is separable from their role as pacemaker neurons: ablation or functional silencing of the LN_vs reduces cocaine sensitivity, while loss of the principal circadian neurotransmitter PDF has no effect. Together, these results reveal a novel role for Lmo in modulating acute cocaine sensitivity and circadian locomotor rhythmicity, and add to growing evidence that these behaviors are regulated by shared molecular mechanisms. The finding that the degree of cocaine responsiveness is controlled by the Drosophila pacemaker neurons provides a neuroanatomical basis for this overlap. We propose that Lmo controls the responsiveness of LN_vs to cocaine, which in turn regulate the flies' behavioral sensitivity to the drug.     

A Cognitive Neural Architecture Able to Learn and Communicate through Natural Language

Communicative interactions involve a kind of procedural knowledge that is used by the human brain for processing verbal and nonverbal inputs and for language production. Although considerable work has been done on modeling human language abilities, it has been difficult to bring them together to a comprehensive tabula rasa system compatible with current knowledge of how verbal information is processed in the brain. This work presents a cognitive system, entirely based on a large-scale neural architecture, which was developed to shed light on the procedural knowledge involved in language elaboration. The main component of this system is the central executive, which is a supervising system that coordinates the other components of the working memory. In our model, the central executive is a neural network that takes as input the neural activation states of the short-term memory and yields as output mental actions, which control the flow of information among the working memory components through neural gating mechanisms. The proposed system is capable of learning to communicate through natural language starting from tabula rasa, without any a priori knowledge of the structure of phrases, meaning of words, role of the different classes of words, only by interacting with a human through a text-based interface, using an open-ended incremental learning process. It is able to learn nouns, verbs, adjectives, pronouns and other word classes, and to use them in expressive language. The model was validated on a corpus of 1587 input sentences, based on literature on early language assessment, at the level of about 4-years old child, and produced 521 output sentences, expressing a broad range of language processing functionalities.