A Multiple Memory Systems Approach to Economic Decision Making

Multiple Memory Systems
Image Unavailable
An illustration of how memory stores are commonly divided in psychology
Source: http://www.studyblue.com/notes/note/n/bio-2232-exam-2/deck/5668604

Much of human choice behavior, especially that in economic situations, is perceived as highly hedonistic and self-motivated, i.e., our decision making process is underpinned by a calculus of what action will elicit the greatest personal gain. For this reason, it is intuitive that many neuroeconomic investigations into the neural circuitry underlying choice behaviors have focused on ‘pleasure’ centers in the brain, as well as those substrates involved in reward-mediated learning and memory. Much of these had been already been identified throughout the 20th century, and the role of areas including the ventral tegmental area, the basal ganglia, and the dopaminergic/cholinergic diffuse modulatory systems in reward-conditioned behavior had been well documented [1]. For this reason, these nuclei and pathways became a focal point for neuroeconomic studies of brain substrates involved in choice behaviour [2]. Through the use of animal models, fMRI studies, and investigations into the effect of various dopaminergic/striatal pathologies (ex. Parkinson's Disease) on decision-making, two things became clear: (1) these brain regions play a role an important role, but (2) cannot fully account for the high degree of complexity and variability surrounding human choice behavior [2].

Upon further investigation, an array of different functional brain regions became implicated in choice behavior, thus complicating our understanding of it. The medial temporal lobes (MTL), sites crucial for the consolidation of emotional and declarative memory, were identified as being a key component in choice behavior [3]. In addition, aspects of the frontal lobes involved in working memory and executive function were also implicated. Other regions have been identified as well, but will not be discussed here [2].

To add further complexity, the role that these different functional regions play in the choice behaviors appears to be far from consistent, varying dramatically with the context of the decision to be made. This finding has lead to current investigations into the nature of the connections between these multiple memory systems, and further, how the nature of the connection is influenced by context [2].

This portion of the ‘Neuroeconomics’ neurowiki will focus on the basic anatomy and physiology of these different neural memory systems and how they relate to choice behavior, with an emphasis put on (1) the basal ganglia, and (2) the medial temporal lobes. In addition, it will introduce the emerging theories pertaining to the relationship between these multiple memory systems, and how this relationship culminates into human choice behavior.

Section 1 Reward mediated Learning and Memory: Basal Ganglia and Dopamine

1.1 Introduction

Reward-mediated conditioning (RMC), which can be crudely interpreted as ‘trial-and-error learning’ is an important learning and memory system involved in choice behavior. It allows you to sort out, given a certain context, which decisions will lead to the desired outcome, and which will not, in an experience based manner [2]. The name ‘reward-mediated conditioning’ is derived from the observation that when an action results in the aquisition of a reward, a subject will be more likely to repeat the action, should a similar situation present itself. In this way, a learned association is formed between a situation, action and outcome [4]. This form of learning is regarded as ‘implicit’, as the resulting association of stimulus-response or action-outcome (i.e. learning) following RMC occurs below the level of consciousness [4].

Over the years, RMC has been extensively investigated and observed in organisms as simple as C elegans and D melanogaster, and as complex as rodents and primates [5]. In humans, RMC is suggested to be important in a number of behavioral functions, especially choice behavior during economic games, where a subject must learn what choice/action, in a given context, will maximize the reward they receive, or minimize punishment. Because such games are commonly used in the study of neureconomics, it is intuitive that the neural mechanisms underlying this form of learning and memory have become an important target for research in this field [2].

1.2a Dopamine and the Midbrain: the Ventral Tegmental Area and Mesocorticolimbic Pathway

Intracranial Self-Stimulation
Image Unavailable
A simplified example of the machinery used to facilitate
intracranial self-stimulation in rats
Source: http://www.cerebromente.org.br/n18/history/stimulation_i.htm

In 1954, Olds and Milner discovered that stimulation of the medial forebrain bundle (MFB) provided a strong positive reinforcement in rats [6]. This is to say, that MFB stimulation could be used as a reward in RMC, and produced behavioral effects consistent with those observed following the administration of rewarding drugs, such as cocaine or amphetamine (i.e., conditioned place-preference, exploratory behaviors, etc.) [7]. When the rats were given control over their own stimulation (via a bar-press lever), the positive effect of the intracranial self-stimulation was often so strong that the animal would continue to bar-press while neglecting food, water and sex [7]. In these cases, it appeared as though the natural reward system of the brain had been effectively over-ridden.

It soon became apparent that stimulation of the ventral tegmental area (VTA) in the mesencephalon could produce similar effects [8]. This finding was consistent with neuroanatomical knowledge; the axonal projections of the VTA are the main constituent of the MFB, i.e., both structures are anatomically connected, and stimulation of either one should produce similar effects. The VTA and MFB are structures common to a diffuse ascending dopaminergic system called the mesocorticolimbic pathway, which originates in the VTA itself [8]. The finding that stimulation of this pathway could produce strong positive reinforcement led to further investigations of diffuse ascending dopaminergic pathways in the brain, and they quickly became regarded as a neural substrates important in the subjective feelings of ‘pleasure’ and ‘reward’ [8].

The mesocorticolimbic limbic pathway, is one of two diffuse ascending dopaminergic pathways in the brain. It can be subdivided into two separate functional systems, (1) the mesolimbic pathway, and (2) the mesocortical pathway. As the names suggest, the projections of the mesolimbic pathway mainly target limbic structures such as the nucleus accumbans, as well as the hippocampus and amygdala, whereas those of the mesocortical pathway mainly target regions of the frontal lobes [8]. The function of the mesocorticolimbic pathway has largely been elucidated by studies involving intracranial stimulation, such as the one mentioned above, or studies using drugs that act of dopaminergic synapses such as amphetamine and cocaine [7]. These drugs, which provide strong positive behavioral reinforcement in humans and animal models, have been shown to increase synaptic DA concentrations in the VTA. In turn, they result in the up-regulation of activity in the mesocortical and mesolimbic pathways, which has been suggested to be the basis of reward-conditioned behaviors, and the subjective feeling of reward [8]. To compliment these findings, several studies have illustrated that, following administration of pharmaceutical agents that antagonize DA action or reduce DA concentrations in the VTA, there is a reduction of subjective feelings of pleasure following reward (in humans), and a decreased capacity for reward-mediated conditioning (in animal models) [9]. For these reasons, these pathways have been implicated in the cognitive and emotional aspects of reward, and are regarded as important in mediating the behavioral changes that occur during RMC [8].

This conclusion was further supported in an interesting study done by Hsing-Chen Tsai et al. (2009) [10] where they employed the use of the optogenetic methods refined by Karl Deisseroth. Using flashes of light at a certain wavelength, the researchers were able to selectively control the firing rate of only DA neurons in the VTA. In this way, different temporal patterns of neuronal firing (i.e., phasic, tonic, etc.) were elicited, and the researchers were able to conclude that phasic firing of VTA DA neurons was sufficient to induce conditioned place preference in mice, a form of RMC. Because these neurons are the basis of the diffuse ascending DA pathways, including the mesocortical and mesolimbic pathways, this study reiterated their importance in mediating RMC.

1.2b Dopamine and the Midbrain: the Substantia Nigra and Striatum

Diffuse Ascending Dopamine Pathways
Image Unavailable
Illustration of the mesocorticolimbic and nigrostriatal dopamine pathways
in the human and rat brain
Source: http://www.news-medical.net/health/What-is-Dopamine.aspx

The second diffuse ascending DA pathway is the nigrostriatal pathway. This pathway originates in the substantia nigra pars compacta (SNc), located in the mesencephalon, and ascends upwards to the striatumstraitum [11]. For many years, it was assumed that the main function of this pathway was to refine movement by providing variable inputs to the motor cortices. This assumption was largely based upon knowledge derived from studies of Parkinson’s Disease (PD), where the selective neurodegeneration of neurons in the SNc is associated with various motor impairments such as bradykinesia and resting tremor [1]. However, in recent years it has become apparent that this pathway, especially components of the striatum, are important for processes outside of coordinated motor movement, such habit formation, skill learning, emotional responses and reward-mediated conditioning [12].

The multi-functional nature of the striatum has been a topic of recent popularity in the study of RMC. A study done by Kawagoe, Takikawa and Hikosaka (1998) [13] showed that, in a RMC test where subjects has to perform an eye-saccade following a conditioned stimulus, the caudate nucleus, a component of the striatum, played an important role in integrating information pertaining to both eye saccades and the expectation of reward. The finding that the striatum is involved in the anticipation and perception of reward has been further supported in many other studies, resulting in a general acceptance of the importance of both the striatum and nigrostriatal pathway in both the subjective feeling of reward and RMC [12]. This conclusion appears to illustrate a neural mechanism of RMC common to rodents, non-human primates and human beings.

1.3a Reward-Mediated Conditioning in Human Subjects: Neuroimaging

Despite this common neural system underlying RMC, some caution must be taken, as the vast majority of the evidence that supports these conclusions is derived from studies of animal models, namely rodents. These animal models (1) contain highly conserved brain structures analogous to those found in human beings (ex. the basal ganglia), and (2) are genotypically similar to human beings. However, some argue that researchers lack justification in asserting that conclusions made about neural mechanisms underlying rat choice behavior, in a regulated laboratory setting, can be generalized to human beings and the complex real world situations they are faced with [7]. Unfortunately, though this argument does carry weight, researchers hands are often tied, as many of the experimental conditions employed with animal models would be highly unethical or unpopular with human subjects [7]. For example, we cannot ethically administer amphetamine and cocaine to human non-addicts, and many would be hesitant to have an electrode implanted into their midbrain or medial forebrain bundle. As such, alternative methods of studying the neural mechanisms underlying RMC, such as neuroimaging and lesion/pathology studies, have become particularly important in studies using human subjects [2].

PET Scan Using Raclopride Model
Image Unavailable
Figure from a study that illustrated striatal increases in dopamine
using a PET scan and the raclopride model [14]

Neuroimaging studies that have made use of technologies such as positron emission tomography (PET) and functional magnetic resonance imaging (fMRI) have allowed researchers to confirm inferences about human choice behavior derived from animal models, and opened doors in understanding phenomena that were previously closed [12]. It has become clear that, like animal models, basal ganglia and dopaminergic pathways are important to human RMC, and appear to behave similarly to the animal models themselves. For example, through the use of a ‘raclopride’ model, PET studies have illustrated that DA concentration in the human VTA and striatum increase during RMC exercises [14]. In this model, carbon-11-radioisotope labeled raclopride(a D2 receptor antagonist) is administered to the subjects prior to the conditioning exercise. Because the raclopride is radiolabelled, its regional concentration can be monitored using PET. When higher levels of DA present, the DA will exert stronger competition with the drug to bind to D2 receptors, and regional concentration (in this case within the VTA/striatum) of raclopride (i.e., radioactivity) will decrease; a change that can be measured by the PET scanner [12][14].

Using creative imaging techniques like this, many studies have illustrated the similarities between the neural mechanisms of RMC in humans and animal models. However, several differences have also been observed. Notably, several fMRI studies have shown a higher degree of activation of the dorsal striatum (i.e., head of the caudate nucleus) in human subjects during RMC, than that observed in animal models, where the activity is predominantly localized to the ventral striatum (i.e., nucleus accumbans) [12]. This structural distinction appears to have functional implication; in cocaine addicts, the subjective sensation of ‘craving’ is associated with activity in the ventral striatum, whereas the sensation of ‘pleasure’ following cocaine administration was associated with activity in the dorsal striatum [15]. Other neuroimaging studies have suggested that the human dorsal striatum also serves to support the capacity to differentiate between positive and negative reinforcement [16].

1.3b Reward-Mediated Conditioning in Human Subjects: Parkinson’s Disease

Finally, further knowledge of the neural mechanisms underlying human RMC comes from studies that examine the behavior of subjects with pre-existing lesions/pathologies relative to healthy controls. Because of the implication of the nigrostriatal pathway in RMC, subjects with Parkinson's Disease (PD), which causes the selective degeneration of DA neurons in the SNc, have been especially targeted in these types of studies [2]. Due to the gradual loss of activity in the nigrostiratal pathway due to SNc neurodegeneration, it is not surprising that subjects with PD often show marked deficits in feedback-based and associative learning paradigms, such a RMC, relative to healthy controls [17]. These deficits are assumed to be associated with an observed decrease in activity in the dorsolateral striatum in response to reward-prediction exercises [18]. Interestingly, the administration of levodopa (a DA precursor that can cross the blood-brain barrier) allows for a significant increase in the capacity for PD subjects to respond to RMC, and increases the subjective feeling of ‘pleasure’ associated with the reward [19].

Importantly, though subjects with PD exhibit deficiencies in RMC and other forms of associative learning, other learning and memory systems are generally not affected [2]. Furthermore, these intact memory systems can often compensate for the large-scale destruction of memory systems mediated by the mesencephalon and basal ganglia. Because of this, PD subjects will often not appear to present any cognitive deficits, and often perform normally in the economic games used in neuroeconomic studies [2][12]. It is in the administration of tests that target learning and memory systems in the BG, that deficits associated with PD are exposed.

1.4 Conclusions

In conclusion, all of the evidence presented above converges on the idea that (1) the mesocorticolimbic and nigrostriatal dopaminergic pathways, (2) their nuclei of origin in the mesencephalon, and (3) the brain regions they project to (i.e., BG and frontal lobes), are all important neural contributors to associative forms of learning and memory, such as RMC. As described at the beginning of this section, it is these forms of learning that allow for an organism to maximize reward and minimize punishment in an experience-based manner. For this reason, it is no surprise that these structures have been implicated as important neural contributors to human choice behavior, especially in the reward-based economic games common to studies in the field of neuroeconomics. Despite their importance, evidence-some of which was presented above-has suggested that they do not act alone in this function, and are rather a single component of an interactive ‘multiple-memory system’ that underlies choice behavior [2]. The other components that constitute this system will be discussed in detail in the following sections.

Section 2 Declarative Memory and Theory of Mind: the Medial Temporal Lobes

2.1 Introduction

Multiple Memory Systems and Choice Behavior
Image Unavailable
A schematic that shows the multiple forms of memory that can inform
choice behaviour and their neural substrates [2]

Section 1, ‘Reward-Mediated Learning and Memory: the Basal Ganglia and Dopamine’ illustrated a learning and memory system that allowed for the presence of a reward to influence choice behavior in both animal models and human beings. Despite the importance of this neurological system in understanding the basis of choice behavior, especially that displayed in economic games, it is clear that it does not work alone. Rather, it appears to be a single component of an interactive ‘multiple-memory system’ that underlies choice behavior [2][20]. Several pieces of evidence support this claim. First, subjects with PD, a disease causing selective degeneration of DA neurons important in RMC, display impaired, but still highly functional choice behavior in various experimental settings [18]. Second, as should be clear based on one’s own introspection, human beings employ a number of different tactics during choice behavior, such as using information derived from our own past experiences, the experiences of others, or attempting to take the perspective of another person [2]. These phenomena cannot be explained by appealing exclusively to the fairly simplistic basal ganglia and dopaminergic systems mentioned in section 1. Rather, as suggested above, multiple memory systems must be taken into account [20]. It is generally accepted that the system underlying the formation of declarative memory, largely associated with activity in the medial temporal lobes (MTL), is one of these [2][4].

2.2 Understanding the Dissociation Between Multiple Memory Systems: H.M.

In contrast to RMC, which is an inflexible, subconscious association of (1) stimulus, (2) response and (3) outcome, declarative memory is both consciously available to the subject and can be applied flexibly to different situations [4]. The theoretical concept of ‘declarative memory’ can be further subdivided into episodic and semantic memory, which are defined as memory stores for one’s autobiographical past and general facts/events, respectively [4]. Studies over the past fifty years have illustrated that the neural substrates associated with declarative memory are largely dissociated from those associated with implicit forms of memory, such as RMC. H.M., a classic subject in studies of learning and memory, was a remarkable source of insight into this dissociation [4][21].

Due to severe epileptic seizures, H.M. had large portions of his MTL removed bilaterally. This operation was successful in reducing the frequency and intensity of his seizures, however, it left him incapable of forming new declarative memories, a condition referred to as ‘anterograde amnesia’. Despite this debilitating affliction, his short-term, working, and non-declarative (ex. RMC) memory systems remained unaffected. For example, during the process of his psychological testing, he got progressively better at complex motor tasks that he was asked to perform in the lab, even though he exhibited no awareness of ever having performed the task before. Another interestingly finding was that his declarative memories from before the operation remained largely intact. The bizarre nature of this symptomatic presentation lead to the suggestion that the regions removed from his brain, i.e., the MTL, contained neural systems vital for the formation of declarative memories, but less important for the both the retrieval of declarative memories and function of other memory systems [21]. Now, the reality of this hypothesis has become clear and has led us to a better understanding of how declarative memories are both formed and recalled at a neurological level [2][4][21].

2.3 Models and Neuroanatomy of Declarative Memory

From an anatomical standpoint, the MTL is composed of a multitude of substructures; including the entorhinal cortex, hippocampus and parahippocampus, just to name a few [1][2]. A recent review by Delgado and Dickerson [2] implicates all of these structures as important neural substrates in functions such as the detection of novel stimuli, the formation of memories for facts and events (i.e., semantic/episodic memory), spatial navigation and mediating the transfer of knowledge between contexts. Studies that have analyzed the effects of lesions in these regions, such as those with H.M., have illustrated that damage to these regions can result in the impairment of any or all of these functions, in both human and non-human subjects [21].

Models of Declarative Memory Consolidation
Image Unavailable
This figure illustrates the difference between the standard consolidation and
multiple trace theories of declarative memory consolidation by showing the projected effects
of hippocampal lesions of different memory stores
Source: http://www.nature.com/nrn/journal/v6/n2/box/nrn1607_BX3.html

Of all the functionally distinct structures in the medial temporal lobes, the hippocampus has arguably been the most extensively analyzed [4]. It is generally accepted that it plays a vital role in the consolidation of information from the short-term to long-term memory stores, however, the precise mechanismis still a topic of debate [2][4]. The ‘standard consolidation theory’ (SCT) suggests that long-term memories (LTM) are stored in the cortex (primarily association cortices) as a network of synaptic connections, which is orchestrated and strengthened by neuronal projections from the hippocampus [22]. Because the hippocampus receives input from sensory and short-term memory stores, it can be seen to be essentially ‘bridging-the-gap’ between the representation of information in short-term stores, i.e., primary sensory cortices and regions of the frontal lobes, and its representation in long-term stores, i.e., synaptic networks spanning association cortices [4]. Importantly, in the SCT, the hippocampus functions to create and enforce a set of cortical synaptic connections, and in this way, is not viewed as either the place of memory storage, nor important to the existence of a memory once it has been consolidated into the cortex [4][22]. This makes sense in reference to H.M., as the removal of his MTL, and as such, his hippocampus, resulted in the loss of ability to form new memories, while memories from before the operation were preserved [21].

Despite the elegance with which the SCT explains the case of H.M., other theories of long-term memory storage suggest that the hippocampus is important for both the formation and retrieval of memories [4]. This suggestion seems incongruent with H.M.’s case, however, Nadel and Moscovitch made an important distinction to alleviate this incongruence. In their ‘multiple memory trace’ theory (MMT), they posit that the SCT only applies to semantic memories, i.e., only semantic memories are stored as cortical networks that become independent from the hippocampus over time [23]. Episodic memories, on the other hand, are suggested to require hippocampal activity both during the process of consolidation and retrieval, and therefore never become fully independent from it [23]. As such, this theory would suggest that H.M. had effectively lost all episodic memories, but still retained semantic memories, or semantic details of episodic memories, from before the operation.

Though these theories are elegant in their application to studies of subjects with lesions/pathologies, they remain poorly defined from a neurological standpoint [4]. For example, the difference between episodic and semantic memories is not well understand at a cellular level, and furthermore, the neurological nature of all LTM remains a controversial topic. As such, these theories should be viewed as broad-spectrum frameworks that can be used to inform a neurological understanding of declarative learning and memory, rather than an explanation of the learning and memory systems themselves. Despite this, there is a wealth of studies that have analyzed the neurophysiological nature of memory consolidation, yielding many interesting results that will be discussed in greater detail below [2][4].

2.4 The Neurobiology of Declarative Memory: Long-Term Potentiation and Depression

Long Term Potentiation and Depression
Image Unavailable
Illustration of the molecular mechanisms underlying LTP/LDP
in a glutamatergic synapse
Source: http://www.nature.com/nrn/journal/v8/n11/fig_tab/nrn2234_F2.html

One of the most important discoveries that occurred in the wake of the studies mentioned above, was that of long-term potentiation (LTP) and long-term depression (LTD), phenomena which are observed at the synaptic level. The terms LTP and LTD describe the observed increase in post-synaptic depolarization (i.e., excitability) following the coupling of pre-synaptic/post-synaptic activity, and the decrease in such post-synaptic depolarization in the absence of said coupling, respectively [1]. Because these opposing processes allow for the nature of the electrical relationship between two neurons to change without the gain/loss of entire synapses or neurons, they are posited to be a fundamental mechanism of plasticity in the brain [24]. In this way, it appears that these mechanisms are a potential means through which the brain ‘learns’. LTP and LTD occur in synapses throughout the nervous system, however, from a molecular standpoint, the mechanisms underlying LTP in glutamatergic (excitatory) synapses have been most extensively studied [1]. Not so coincidentally, the hippocampus is a brain region that has a large presence of such synapses.

Currently, it is thought that the simultaneous binding of glutamate to post-synaptic NMDA(glutamatergic) receptors and depolarization of the post-synaptic membrane, allows for an influx of calcium ions into the post-synaptic neuron via the NMDA channel [1][25]. Within the cell, the calcium ions exert their effects by activating a number of kinases, of which calmodulin kinase II alpha is particularly important. In turn, this allows for the kinase-dependent movement of AMPA receptors (glutamatergic) to the post-synaptic density, as well as the phosphorylation of AMPA receptors that were already present. In this way, the increased presence and activity of AMPA receptors causes subsequent release of glutamate from the pre-synaptic neuron will result in a greater post-synaptic depolarization [24]. It is thought that LTD works in the exact opposite fashion, i.e., AMPA receptors are de-phosphorylated and removed from the synaptic membrane by a set of phosphatases, namely calcineurin [1].

Due to the high presence of these excitatory synapses in the hippomcapus, this form of LTP/LTD has been extensively studied in this region. In particular, it has been shown that the modification of synapses between the axons of CA3 neurons (Schaffer Collateral) and the dendrites of CA1 neurons occurs via the molecular mechanisms describe above [26].

As alluded to above, it appears as though the synaptic changes associated with LTP and LTD are a potential means by which the brain ‘learns’. Due to the high concentration of glutamatergic synapses in the hippocampus, it seems reasonable to suggest that these mechanisms are important to the formation of declarative memories [4][24]. This hypothesis has proven to be largely correct, as disruptions of LTP via the administration of NMDA antagonists has been shown to result in the impaired consolidation of declarative memories [1]. Note, however, that this does not show what a declarative memory actually consists of, but rather, illustrates a mechanism that assists in its formation. The molecular changes that follow LTP/LTD in the post-synaptic neuron are complex, and how these changes constitute a ‘memory’ remains a topic of heated debate.

With a better understanding of the neurological basis of declarative memory consolidation, we can now step back to analyze how this form of learning and memory pertains to choice behavior. As mentioned at the beginning of this section, human beings rely on our episodic and semantic memories in the decision making process. For example, when deciding whether or not to eat at a particular restaurant, one might consider past experiences with the type of food served there, or reviews they have heard from friends or media sources. Therefore, because the molecular and neurological systems (i.e., LTP/LDP and MTL) presented above support the consolidation of declarative memories, it follow that they play an important role in choice behavior as well.

2.5 Medial Temporal Lobe Involvement in Other Aspects of Choice Behaviour

Common Neural Pathway
Image Unavailable
fMRI representation of the common neural system underlying declarative memory,
prospection and theory of mind as proposed by Buckner & Carroll [3]

One study by Shohamy et al. (2008) [27] indicates that the MTL are important to another aspect of human choice behavior; the reversal or modification of learned associations. In this study, amnesiac subjects with bilateral hippocampal damage were exposed to a probabilistic learning paradigm requiring an association to be formed between stimulus, action and outcome, much like RMC (see section 1). Relative to healthy subjects, those with hippocampal damage were relatively un-impaired in their capacity to learn the stimulus-response pairing, a capacity more dependent on the integrity of the basal ganglia. However, when subjects were told to change a response to a stimulus that had already been paired to another response (a process called reversal), amnesic subjects were largely incapable of doing so. In this way, the researchers concluded that MTL structures somehow serve to reverse, or modify, learned associations. This has a clear implication in choice behavior, as in order to effectively achieve a desired end, one must retain the capacity to shift learned associations in order to conform to a changing external environment.

Finally, another important aspect of human choice behavior supported by the function of the MTL, especially in social situations, is the capacity to employ ‘theory of mind’ (TOM), i.e., to put oneself in another’s ‘shoes’ [2]. In doing so, we can attempt to understand the motivations and strategies of others, in order to modify our own behaviors and decisions in an appropriate manner. In a review by Buckner and Carroll (2007) [3], it is argued that there is converging evidence on a common neural pathway important to both the formation of declarative memories and utilization of TOM. Most of the studies cited in this review are fMRI analyses that show activation in frontal lobes, MTL, and medial parietal lobes when subjects form/retrieve declarative memories or employ TOM. In addition, the researchers suggest that many of the structures in this common pathway are also important for mentally simulating future situations (prospection), another cognitive function utilized during human choice behavior.

2.6 Conclusions

It should now be clear that research from different fields has converged on the importance of the MTL in various cognitive functions involved in human choice behavior. These include (1) the consolidation and retrieval of declarative memories, (2) the utilization of theory of mind, and (3) the capacity for prospection. However, it would be naïve to assume that these are the only functions that the MTL support. In fact, the MTL have been implicated in a number of different emotional and memory circuits in the brain, and the anatomical connection of the hippocampus to aspects of the basal ganglia, Papez circuit, and frontal lobes, suggests its importance in a number of different brain functions [1][3][4]. Recently, it has been suggested that connections between the hippocampus, striatum and ascending DA pathways, allow for MTL memory systems to become involved in RMC [2]. In this way, the multiple memory systems involved in choice behavior may not be as dissociated as originally thought; an idea which will be discussed in great detail in Section 4: ‘Understanding interactions of multiple memory systems in economic decision making’.

Section 3 Executive and Alternative Learning and Memory Systems

3.1 Introduction

The past two sections have outlined learning and memory systems responsible for the subconscious association of action, outcome and outcome value (i.e., reward-mediated conditioning [RMC]), as well as for the consolidation of memories into longer-term storage (i.e., declarative memory). In addition, facets of the latter system have been implicated in other important aspect of human choice behavior, such as theory of mind and prospection. Though there is much evidence that supports a neurological dissociation between these memory systems, i.e., between the medial temporal lobes (MTL) and the basal ganglia (BG), there is an increasing body of evidence that suggests that this dissociation may not be as absolute as once thought. This claim is supported by neuroanatomical insights that illustrate the structural and functional interconnectedness between these memory systems, and between each respective system and other regions of the brain [1]. Recently, the frontal lobes and amygdala have become regarded as important regulators and modulators of the connection between these two memory systems, as well as various aspects within each of them [2][28]. Furthermore, they have also been suggested to contain neurological substrates that support other memory systems important for human choice behavior [4]. This section will outline the structure and function of the amygdala and frontal lobes, and provide insight into their relationship to choice behaviors exhibited by both humans and animal models.

3.2 The Amygdala and Choice Behavior

Connections of the Amygdala
Image Unavailable
A neuroanatomical representation of the axonal projections of the amygdala,
some of which are vital in mediating fear-conditioned responses
Source: http://www.neilslade.com/chart.html

Located at the tail of the caudate nucleus, the amygdala lies deep in the MTL proximal to the hippocampus [1]. Despite its spatial location in the MTL, it was not regarded in the previous section, as its function and connections are relatively distinct from the declarative memory circuitry associated with the hippocampus [20]. Despite this functional distinctiveness, there is still a large degree of communication between the hippocampus and the amygdala, so the two structures should not be considered entirely dissociated from one another. Generally, the amygdala is considered an important neural contributor to emotional learning, especially when aversive stimuli or events are involved [28].

Namely, it mediates the formation of strong associative memories between a particular stimulus and an undesirable outcome (fear conditioning), allowing for the conditioned stimulus to produce a strong fear response in subsequent presentations, even if the aversive outcome is itself not present [1][4]. This fear response is thought to be the result of communication of the amygdala with various brain regions, including the hypothalamus, cortical regions and midbrain nuclei. A particularly important target in the midbrain is the periaqueductal grey (PAG), which mediates fight, flight and freezing behaviors, all of which can be characteristics of a fear response [29].

The capacity for the amygdala to mediate emotional learning makes sense, given its structural location between the nucleus accumbans, an important ‘emotional’ nucleus in the Papez Circuit, and the hippocampus, and important region for the formation of declarative and spatial memories [29]. Furthermore, this regional positioning of the amygdala, as well as its connectivity to other brain regions, serves its capacity to provide input into both striatal and hippocampal structures to their activity [30]. This input has caused the amygdala to be implicated in both RMC and the formation of spatial and declarative memories in the MTL [2]. In terms of RMC, connections between the basolateral amygdala, midbrain nuclei and the nucleus accumbans have been shown to be vital in the formation of appropriate action-outcome associations [28]. These connections become increasingly important when the action is being coupled with more aversive outcomes. Interestingly, fMRI studies in which subjects are faced with an ambiguous decision have shown increased blood-oxygen-level dependent (BOLD) responses, i.e., greater activity, in the amygdala, orbitofrontal cortex and insular cortex [31]. This finding is intuitive given that human beings generally do not like-and exhibit aversive behaviors towards-situations where a decision must be made with ambiguous contextual information [2]. This finding is also important as it implicates the amygdala as an important neural substrate involved in choice behaviors during uncertain situations [28].

Due to the capacity of the amygdala to modify activity in midbrain dopaminergic structures, the striatum, the nucleus accumbans and the hippocampus, it has been suggested to play an important role in mediating the interaction between the memory systems supported by the BG and MTL [2][28]. Because both of these memory systems have been shown to play important role in choice behavior, it follows that any structure that modifies their activity and interactions, i.e., the amygdala, can be seen as important in choice behavior as well [28]. In summary, the amygdala’s role in fear conditioning, decisions in ambiguous contexts, and the modulation of activity in BG and MTL memory systems, suggests that it plays an important role in human choice behavior.

3.3 The Frontal Lobes and Choice Behavior

Structure and Divisions of the Frontal Lobes
Image Unavailable
A neuroanatomical representation of the various regions of the frontal
and pre-frontal cortices, some of which play a role in choice behaviour
Source: http://mindblog.dericbownds.net/2008/05/models-of-cognitive-control-in.html

The frontal lobes, especially aspects of the prefrontal cortex (PFC), represent a second source of modulatory input into the BG and MTL memory systems [28]. Please note that the breadth of research into the effects of the frontal regions on learning and memory is vast, and therefore only a brief review of important findings will be discussed here. For an extensive review on the frontal lobes and their effect on learning and memory and choice behavior please see the article by Rangel, Camerer and Montague (2008) [28].

To start, the orbitofrontal cortex (OFC) and amygdala have been shown to play mutually supportive roles in a variety of functions, such as (1) learning how conditioned valence-neutral stimuli can predict the value of an associated outcome [32], and (2) mediating decision making in ambiguous or uncertain settings [28]. The OFC also appears to be an important neural center for the encoding of the value of outcomes. This is illustrated in fMRI studies that show that the OFC, along with the dorsolateral and ventromedial PFC (dlPFC and vmPFC, respectively), present increased BOLD responses during planning for positive goal-directed behaviors [33]. The OFC is generally seen as an important aspect of RMC, as it compliments that activity of the striatum, which forms action-outcome associations, by mediating the formation of outcome-value associations [28]. In this way, the OFC and striatum work together in forming the action-outcome-value pairing that is the foundation of RMC [12].

Interestingly, the human striatum has been shown to not only associate with the OFC, but also with the vmPFC, vlPFC (ventrolateral), blPFC, and several midbrain DA structures [28]. The set of these connections forms a ‘striatal-prefrontal network’, suggested to be important in mediating a number of functions. First, this network has been observed to become active during goal-directed planning exercises, especially those in uncertain and ambiguous situations [33]. Second, it appears to support the capacity to ‘weight-out’ the value of short vs. long-term goals [28]. This second function is of particular neuroeconomic importance, as many studies employ the use of games where a choice must be made between a short-term and a long-term reward. Third, increased activity of this network often corresponds to the suppression of activity in the BG and MTL memory systems, particularly during probabilistic learning exercises [34]. This finding lead to the hypothesis that the PFC regions involved in this network (i.e., vm/vl/dmPFC) are important for suppressing/regulating the activity of the more primitive BG and MTL memory systems, in order to allow for more complex goal-directed behaviors [28][31][34]. The nature of the relationship between the PFC and these multiple memory systems will be discussed in more detail in Section 4: ‘Understanding interactions of multiple memory systems in economic decision making’.

Striatal-Prefrontal Network
Image Unavailable
A neuroanatomical representation the strital-prefontal network associated with
several different aspects of choice behavior
Source: http://www.impulsecontroldisorders.org/html/cravings.html 

Finally, regions of the PFC are also suggested to contain the neural substrates that support working memory (WM), yet another distinct learning and memory system [35]. The concept of ‘working memory’, originally proposed by Baddeley and Hitch in 1974, is a theoretical model of the short-term memory processors in the brain [4]. In this theory of WM, there are two short term memory stores; the phonological loop (auditory information) and the visuo-spatial sketchpad (visual information). Controlling these ‘slave systems’ is the central executive, which mediates the movement of information into and out of the two memory stores, as well as the manipulation of information within them [4]. Due to the capacity for WM to hold and manipulate information coming from the primary sensory cortices, as well as long-term memory stores, some researchers argue that it is the basis of human consciousness [4].

Function of the central executive, arguably the most important component of WM, has been associated with the bilateral activation of the PFC, particularly the dlPFC [35]. This is to say, the human capacity to manipulate information (ex. Mental rotation of a 3D object), seems to be somehow supported by activity in the pre-frontal cortices. In an fMRI study done by Collette and Van der Linden (2002) [35], it was found that there was also significant activation of the left superior parietal lobe during a task that tested the central executive. A number of other studies have noticed the same thing, and as such, it has been suggested that the central executive may constitute a neural circuit that runs between the PFC and medial/lateral parietal lobes [35]. This finding is interesting in its apparent correlation to the circuit common to consolidation of declarative memory, theory of mind, and prospection, described in Section 2.

Though working memory is theoretical model, it has been shown to be an elegant framework for describing many aspects of how sensory, short-term and long-term memory stores can be integrated to allow for a conscious perception of the external environment. Furthermore, the central executive by definition allows for information to be manipulated and applied flexibly to different contexts [4]. The human capacity for consciousness, the manipulation of information, and the flexible application of information to different processes are all important functions employed during choice behavior [2]. Therefore, in the capacity of the working memory model to eloquently provide a framework for understanding these processes, it is likely a useful tool in understanding various aspects of choice behavior as well.

Section 4 Understanding Interactions of Multiple Memory Systems in Economic Decision Making

4.1 Introduction

Who Do You Trust?
Image Unavailable
Source: http://www.savagechickens.com/2005/08/trust-and-respect.html 

As illustrated in the preceding sections, there are multiple memory systems that play vital roles in supporting different aspects of choice behavior. Though there are large-scale dissociations between these multiple memory systems, many studies have illustrated that there is also fairly extensive communication between them [2]. Furthermore, the way in which they communicate appears to be an important factor in modulating different aspects of choice behavior. As such, it has become clear that choice behavior in neuroeconomics cannot be understood by focusing solely on the neurology of the individual memory systems, but must also integrate a knowledge of how these systems interact with each other [2]. Because the BG and MTL memory systems are focal points in understanding the choice behaviors studied in neuroeconomics, the interactions between specifically these two memory systems has been the most extensively researched. Therefore, though the frontal lobes and amygdala appear to support distinct memory systems import to choice behavior (i.e., working and emotional memory), the interactive nature of these systems are not discussed in much detail [2][28]. Rather, these brain regions are generally regarded as accessory structures that are more important for the modulation of activity in the BG and MTL memory systems [28].

4.2 Competitive Memory Systems

One prominent theory that has emerged in the field of learning and memory it that the BG and MTL memory systems may compete with each other in order to facilitate different forms of learning, or an adaptation to changing external contexts [36]. In order to illustrate this phenomenon, many studies make use of a classic economic trust game [2]. This game, which involves the use of two subjects, proceeds as follows: subject A is given a sum of money, say 20$, and is told that they can either keep for themselves or give some amount (or all) of in the other subject (subject B). They are also informed that any money that they give to subject B will be immediately tripled, in this case producing a potential maximum of 60$, and subject B will then have the opportunity to give any amount of that money back to subject A, or keep it all for themselves. The money given back to subject A from B will not be tripled, and the game will be over following this transaction. In this scenario, should both subjects be fully trusting of each other and fair, there would be the potential for both of them to receive 30$, i.e., A gives B 20$ and B gives half of their money (30$) back to A. This is a net gain of 10$ for subject A, considering they could have kept the initial 20$, and therefore seems like the best option. However, it requires that both subjects be trusting and fair [2].

Trust and Choice Behavior
Image Unavailable
An illustration of how good, bad and neutral declarative memories of moral
character effect trust and subsequent choice behaviour [36

Should the subjects not know each other prior to the experiment, the formation of this trust and fairness, i.e., an understanding of how trustworthy one’s partner is, would require probabilistic learning of by trial-and-error [36]. Therefore, it is not surprising that fMRI studies have shown this game to result in significant activation of the basal ganglia, particularly the striatum; structures important to implicit and associative forms of learning [37]. A recent fMRI study done by Delgado, Frank and Phelps (2005) [36], added a slight variation to this game in order to analyze interactions between these associative learning systems and those involved in declarative memory. In this study, subjects were informed of their partner’s moral character prior to playing the game, thus instilling knowledge of how trustworthy one’s partner was in the form of a declarative memory. The researchers noticed that striatal activation was largest in those subjects who were told that their partner was morally ‘neutral’, while activation was close to baseline in those subjects who were told that their partner was morally ‘good’ or ‘bad’. This means that the capacity to learn how trusting one should be of another person is largely inhibited when prior information of that person exists. In other words, this finding suggests that the existence of prior declarative memories results in the suppression of action-outcome conditioning systems in the basal ganglia. In this way, the MTL memory system could be seen as competing with, or suppressing, the BG memory system. This finding has been replicated in a number of contexts, providing further evidence of a competitive relationship between these two memory systems [2].

4.3 Modulatory Activity of the Pre-Frontal Cortex

Following the information presented above, one study noted that the observed decrease of activity of the striatum due to the presence of prior declarative memories was correlated with an increase in activity in the dorsolateral pre-frontal cortex (dlPFC) [38]. This finding is consistent with several studies presented in section 3, which have illustrated the importance of the PFC in mediating the interactions between BG and MTL memory systems [28][31][32][33]. In addition to these findings, the dlPFC has also been shown to interact with the ventral tegmental area (VTA) and nucleus accumbans during RMC, suggesting the importance in the PFC not only in mediating the interactions between BG and MTL memory systems, but also within the memory systems themselves [2]. A recent study by Ballard et al. (2011) [34] suggested that the dlPFC may provide excitatory input to the VTA, in turn activating the diffuse ascending dopaminergic pathways associated with RMC (i.e., mesocorticolimbic pathway). Using fMRI, the researchers noticed a preliminary activation of the dlPFC, prior to the activation of regions of the mesencephalon/BG, during an RMC exercise. This lead them to suggest that the dlPFC serves to integrate information pertaining to the reward, and subsequently passes this information on to the to the VTA and nucleus accumbans (a component of the ventral striatum) via axonal projections. This notion differs-but is not incompatible with-the ideas put forward in section 1, where the VTA was suggested as being the beginning of the neural pathway underlying RMC.

4.4 Modulatory Activity of the Medial Temporal Lobes

HIppocampus (CA3) to VTA Pathway
Image Unavailable
A schematic of the CA3-VTA neural pathway elucidated by Luo et al. (2011) [39]

In addition to the PFC, the hippocampus, an important component of the MTL declarative memory system (see section 2), also appears to play a role in associative forms of learning such as RMC [2]. The precise role it plays in RMC is still far from understood, however, it is thought to be important for the association of context with reward [2][39]. As such, it is likely to play a particularly important role conditioned place preference (a type of RMC). This idea was supported in a recent study by Luo et al. (2011) [39], who elucidated the existence of a neural pathway between the CA3 region of the hippocampus and dopamine (DA) neurons in the VTA.

In this study, the researchers illustrated how the stimulation of neurons in the CA3 region resulted in the subsequent activation of DA neurons in the VTA, in turn causing increased activity of the mesocorticolimbic pathway discussed in section 1. They noticed that the administration of GABA receptor antagonists to the VTA resulted in the absence of this correlated activity with CA3 neurons. As such, it was suggested that the CA3 neurons appeared to activated those in the VTA via ‘disinhibition’, that is, DA neurons in the VTA are chronically inhibited by GABAergic neurons in the caudodorsal lateral septum, and indirect input to this region from CA3 removes this inhibition. To summarize, activity in the hippocampal CA3 region resulted in the indirect activation of the mesocorticolimbic pathway via DA neurons in the VTA. Due to the importance of the mesocorticolimbic pathway in RMC (see section 1) and of the hippocampus in the formation of spatial/declarative memories, the researchers in this study suggested that this pathway is a potential route by which the brain mediates an association between reward and context.

Finally, there is further evidence that suggests that the communication between DA neurons in the VTA and neurons in the hippocampus is bidirectional [2]. This is of functional importance, as some studies have suggested that projections from the diffuse ascending dopaminergic pathways (such as the mesocorticolimbic pathway) to both the hippocampus and frontal lobes support the process of novelty detection [40]. Novelty detection allows for subjects to selectively pay attention to important aspects of the environment, and subsequently consolidate any significant novel information into long-term declarative memory stores [4]. Because the process is vital for forming a perspective on a given context, it is clearly an important precursor to choice behavior within the context itself. As such, the nature and interactions of the diffuse ascending dopaminergic pathways can be seen as important in the mediation of choice behavior due to their role in both RMC (section 1) and novelty detection [2].

1. Bear M.F., Connors B.W., Paradiso M.A. (2007) Neuroscience: Exploring the Brain. 3rd ed. Baltimore (MD): Lippincott Williams & Wilkins.
2. Delgado, M.R., Dickerson, K.C. (2012) Reward-Related Learning via Multiple Memory Systems. Biol Psychiatry 72(2), 134-41.
3. Buckner, R.L., Carroll, D.C. (2007) Self-projection and the Brain. Trends Cogn Sci. 11(2), 49-57.
4. Radavansky G. (2011) Human Memory. 2nd ed. Boston (MA): Pearson Education, Inc.
5. Darvas M, Wunsch AM, Gibbs JT, Palmiter RD. (2014) Dopamine dependency for acquisition and performance of Pavlovian conditioned response. Proc Natl Acad Sci U S A. 111(7):2764-9.
6. Olds J, Milner P. (1954) Positive Reinforcement Produced by Electrical Stimulation of Septal Area and Other Regions of the Rat Brain. J Comp Physiol Psych. 47(6):419-427.
7. Balster RL, Walsh SL, Bigelow GE. (2009) Reflections on the Past 40 Years of Behavioral Pharmacology Research on Problems of Drug Abuse. J Drug Issues. 39(1):133-151.
8. Fibiger HC, Phillips AG. (1988) Mesocorticolimbic Dopamine Systems and Reward. Ann NY Acad Sci. 537:206-215.
9. Yeomans J, Baptista M. (1997) Both nicotinic and muscarinic receptors in ventral tegmental area contribute to brain-stimulation reward. Pharmacol Biochem Behav. 57(4):915-21.
10. Tsai HC, Zhang F, Adamantidis A, Stuber GD, Bonci A, de Lecea L, Deisseroth K. (2009) Phasic firing in dopaminergic neurons is sufficient for behavioral conditioning. Science. 324(5930):1080-4.
11. Wise R.A., Rompre P.P. (1989) Brain Dopamine and Reward. Annual Review of Psychology. 40: 191-225.
12. Delgado MR. (2007) Reward-related responses in the human striatum. Ann N Y Acad Sci. 1104:70-88.
13. Kawagoe R, Takikawa Y, Hikosaka O. (1998) Expectation of reward modulates cognitive signals in the basal ganglia. Nat Neurosci. 1(5):411-6.
14. Koepp MJ, Gunn RN, Lawrence AD, Cunningham VJ, Dagher A, Jones T, Brooks DJ, Bench CJ, Grasby PM. (1998) Evidence for striatal dopamine release during a video game. Nature. 393(6682):266-8.
15. Breiter HC, Gollub RL, Weisskoff RM, Kennedy DN, Makris N, Berke JD, … Hyman SE. (1997) Acute effects of cocaine on human brain activity and emotion. Neuron. 19(3):591-611.
16. Seger CA, Cincotta CM. (2005) The roles of the caudate nucleus in human classification learning. J Neurosci. 25(11):2941-51.
17. Myers CE, Shohamy D, Gluck MA, Grossman S, Kluger A, Ferris S, … Schwartz R. (2003) Dissociating hippocampal versus basal ganglia contributions to learning and transfer. J Cogn Neurosci. 15(2):185-93.
18. Schonberg T, O'Doherty JP, Joel D, Inzelberg R, Segev Y, Daw ND. (2010) Selective impairment of prediction error signaling in human dorsolateral but not ventral striatum in Parkinson's disease patients: evidence from a model-based fMRI study. Neuroimage. 49(1):772-81.
19. Frank MJ, Seeberger LC, O'reilly RC. (2004) By carrot or by stick: cognitive reinforcement learning in parkinsonism. Science. 306(5703):1940-3.
20. Dickerson KC, Li J, Delgado MR. (2011) Parallel contributions of distinct human memory systems during probabilistic learning. Neuroimage. 55(1):266-76.
21. Eichenbaum H. (2013) What H.M. taught us. J Cogn Neurosci. 25(1):14-21.
22. Nadel L, Moscovitch M. (1997) Memory consolidation, retrograde amnesia and the hippocampal complex. Curr Opin Neurobiol. 7(2):217-27.
23. Nadel L, Samsonovich A, Ryan L, Moscovitch M. (2000) Multiple trace theory of human memory: computational, neuroimaging, and neuropsychological results. Hippocampus. 10(4):352-68.
24. Bliss TV, Collingridge GL. (1993) A synaptic model of memory: long-term potentiation in the hippocampus. Nature. 361(6407):31-9.
25. Blitzer RD. (2005) Teaching resources. Long-term potentiation: mechanisms of induction and maintenance. Sci STKE. 2005(309):tr26.
26. Lee HK, Kirkwood A. (2011) AMPA receptor regulation during synaptic plasticity in hippocampus and neocortex. Semin Cell Dev Biol. 22(5):514-20.
27. Shohamy D, Myers CE, Hopkins RO, Sage J, Gluck MA. (2009) Distinct hippocampal and basal ganglia contributions to probabilistic learning and reversal. J Cogn Neurosci. 21(9):1821-33.
28. Rangel A, Camerer C, Montague PR. (2008) A framework for studying the neurobiology of value-based decision making. Nat Rev Neurosci. 9(7):545-56.
29. Phelps EA, LeDoux JE. (2005) Contributions of the amygdala to emotion processing: from animal models to human behavior. Neuron. 48(2):175-87.
30. Phillips RG, LeDoux JE. (1992) Differential contribution of amygdala and hippocampus to cued and contextual fear conditioning. Behav Neurosci. 106(2):274-85.
31. Cardinal RN, Parkinson JA, Hall J, Everitt BJ. (2002) Emotion and motivation: the role of the amygdala, ventral striatum, and prefrontal cortex. Neurosci Biobehav Rev. 26(3):321-52.
32. Balleine BW. (2005) Neural bases of food-seeking: affect, arousal and reward in corticostriatolimbic circuits. Physiol Behav. 86(5):717-30.
33. Tom SM, Fox CR, Trepel C, Poldrack RA. (2007) The neural basis of loss aversion in decision-making under risk. Science. 315(5811):515-8.
34. Ballard IC, Murty VP, Carter RM, MacInnes JJ, Huettel SA, Adcock RA. (2011) Dorsolateral prefrontal cortex drives mesolimbic dopaminergic regions to initiate motivated behavior. J Neurosci. 31(28):10340-6.
35. Collette F, Van der Linden M. (2002) Brain imaging of the central executive component of working memory. Neurosci Biobehav Rev. 26(2):105-25.
36. Delgado MR, Frank RH, Phelps EA. (2005) Perceptions of moral character modulate the neural systems of reward during the trust game. Nat Neurosci. 8(11):1611-8.
37. Poldrack RA, Rodriguez P. (2004) How do memory systems interact? Evidence from human classification learning. Neurobiol Learn Mem. 82(3):324-32.
38. Li J, Delgado MR, Phelps EA. (2011) How instructed knowledge modulates the neural systems of reward learning. Proc Natl Acad Sci U S A. 108(1):55-60.
39. Luo, A.H., Tahsili-Fahadan, P., Wise, R.A., Lupica, C.R., Aston-Jones, G. (2011) Linking Context with Reward: A Functional Circuit from Hippocampal CA3 to Ventral Tegmental Area. Science 333:353-356.
40. Lisman JE, Grace AA. (2005) The hippocampal-VTA loop: controlling the entry of information into long-term memory. Neuron. 46(5):703-13.

Add a New Comment
Unless otherwise stated, the content of this page is licensed under Creative Commons Attribution-ShareAlike 3.0 License