I. Introduction
Every day, we perform complex actions, from speaking a language to navigating busy streets, often without conscious awareness of how we learned these skills. This remarkable ability to acquire knowledge and skills unconsciously is known as Implicit Learning. Closely related concepts include skill learning, procedural learning, and implicit memory (repetition priming). While the brain circuits involved in implicit learning are becoming clearer, the specific neurotransmitters and their functional roles within these circuits remain largely mysterious.
Implicit learning is defined as “the process whereby a complex, rule-governed knowledge base is acquired, largely without any requirements of awareness of either the process or the product of acquisition” [3]. Key characteristics of implicit learning include:
- Limited Awareness: Lack of conscious access to the learned knowledge and how it’s applied.
- Complexity: Acquired knowledge goes beyond simple associations or frequency counts.
- Incidental Acquisition: Learning occurs as a byproduct of information processing, not through deliberate hypothesis testing.
- Non-Declarative Memory: It doesn’t depend on declarative memory systems.
Research indicates that the frontal cortex and basal ganglia, forming fronto-striatal circuits, are crucial for implicit learning in humans. Studies on patients with lesions and neuroimaging of healthy individuals support this [2, 4, 5–7]. The prefrontal cortex is also implicated [4, 5, 8, 9]. Importantly, implicit learning remains intact in amnesic patients, highlighting its independence from the medial temporal lobe (MTL) memory system [4, 10, 11]. In fact, recent research shows MTL deactivation during implicit learning tasks like artificial grammar learning (AGL) [12], reinforcing this distinction. However, a detailed understanding of the learning mechanisms, the nature of acquired knowledge, its representation, and the neural infrastructure – particularly neurotransmitter systems – is still lacking [13].
This article reviews the neuropharmacology literature on implicit learning to shed light on the neurotransmitter systems involved. We focus on neuromodulators – dopamine, serotonin, noradrenaline, and acetylcholine – known for their slow, widespread effects on synaptic transmission. We also consider the roles of γ-aminobutyric acid (GABA) and allosteric α-amino-3-hydroxy-5-methyl-4-isoxazolepropionic acid (AMPA) receptor modulators (ampakines) due to their potential relevance. We also touch upon studies of procedural and probabilistic learning where they offer relevant insights.
The artificial grammar learning (AGL) paradigm is a cornerstone of implicit learning research [14, 15]. It involves an acquisition phase where participants are exposed to sequences generated by complex grammatical rules, often in a short-term memory task. In the test phase, they classify new sequences as grammatical or not, relying on intuition or “gut feeling.” Consistently, participants perform significantly above chance in this task [5, 9, 13, 15], with performance improving over days of implicit acquisition [16, 17].
Another widely used paradigm is the serial reaction time task (SRTT) [15, 18]. Implicit learning is demonstrated by faster reaction times to repeating sequences compared to random sequences. For instance, in a sequence of digits 1-4, each corresponding to a button press, learning the sequence pattern 3-2-4-1 leads to faster responses than random sequences. Participants usually report minimal awareness of the sequence knowledge. While adapted for rodent research, the validity of implicit learning in animals remains debated [19]. Sequence learning, whether in SRTT or AGL, can be seen as a specific instance of learning broader structural regularities or temporal dependencies in stimuli [2, 14].
Tulving categorizes procedural learning (encompassing motor and cognitive skills, conditioning, and associative memories) as being acquired through implicit processes [20]. We will consider tasks often categorized as procedural learning, including puzzles like Tower of Hanoi, London, or Toronto [21], mirror reading/drawing, and tracking tasks. These involve learning complex sequential behaviors. While sequential learning is key in implicit tasks like AGL and SRTT, it’s not always central to all procedural learning tasks. It’s important to note that procedural learning tasks can differ from implicit learning tasks depending on administration. Explicit problem-solving strategies, especially early on, can influence procedural learning. Tasks like mirror reading and trail tracking might not always involve learning novel structured stimuli [4, 15].
Traditional views of procedural learning or skill acquisition emphasize a shift from controlled, explicit processing to automatic, implicit processing as skills develop [22]. Parallel explicit and implicit processes may be involved [4, 13, 23]. Procedural learning tasks usually lack distinct acquisition and test phases, unlike implicit learning tasks. However, neuropsychological studies reveal similar behavioral patterns and neural underpinnings for both, potentially due to a shared implicit component [21, 24, 25]. To ensure robust neuropharmacological findings, converging evidence across various tasks is crucial. Therefore, we will discuss neuropharmacological models of automatization and studies where the implicit component of procedural learning is relevant, acknowledging the potential influence of explicit components.
This review also includes probabilistic learning tasks if they meet the criteria for implicit learning [4, 14, 15], even if the acquired information is less complex. The simplest form is the 2-choice probabilistic learning task, predicting event A or B after a signal, with events occurring probabilistically (e.g., A 75%, B 25%). Probabilities may change, leading to reversal learning. Probabilistic sequence learning is considered implicit [26]. While simpler probabilistic tasks like gambling may involve explicit strategies, their implicit component is still relevant to this review.
“Implicit memory” is often used interchangeably with repetition priming or as a broader term encompassing implicit learning [4]. Implicit memory tasks often test acquisition using exemplars from the learning set (priming), whereas implicit learning often uses novel stimuli based on learned principles [4]. We generally exclude priming studies but include one implicit memory task [27] that aligns with implicit learning criteria and involves generation rather than simple recognition.
This review focuses on research in healthy human subjects but includes animal data for mechanistic insights. Animal studies sometimes use “procedural learning,” but its meaning in animal models can be ambiguous. However, recent rule learning studies in tamarins [28, 29], songbirds [30, 31], and rats [32] are relevant for future neuropharmacological implicit learning research in animals.
Clinical data is considered when other data is lacking or provides unique insights. Related reviews cover general neuropharmacology of cognition and memory [33], working memory [34], basal ganglia in habit formation [35–37], neuropeptides (vasopressin) in animal learning and memory [38], and dopamine in actions and habits [39]. However, a systematic review focused on the neuropharmacology of implicit learning is still missing.
It’s crucial to remember that discussing single effects of pharmacological agents is a simplification. True effects depend on agent concentration, pharmacokinetics, and pharmacodynamics. Dose-response curve studies are more informative but rare in implicit/procedural learning research. Ideally, we should discuss moderate neurotransmitter increases/decreases within physiological ranges. This review tentatively assumes studies are appropriately designed, unless conflicting information arises. However, full dose-response studies are generally lacking.
II. Neuroanatomical Foundations
A brief overview of relevant neuroanatomy, particularly the four main neuromodulators discussed, is provided below.
The dopamine (DA) system has four major pathways (Fig. 1). The mesolimbic and mesocortical pathways originate in the ventral tegmental area (VTA) and project to limbic areas (nucleus accumbens/ventral striatum) and the prefrontal cortex, respectively. Overactivity is linked to euphoria, psychosis, and schizophrenia. The nigrostriatal DA system originates in the substantia nigra and projects to the dorsal striatum; underactivity is associated with Parkinson’s disease. (The tubero-infundibular DA system is not discussed here.)
Fig. 1.
Dopamine pathways (black) and serotonin pathways (grey). Dopamine system significantly overlaps with fronto-striatal circuits important for implicit learning. Serotonin system also shows substantial overlap.
The monoaminergic brainstem, including the raphe nucleus and locus coeruleus, is the origin of the serotoninergic (Fig. 1) and noradrenergic systems (Fig. 2) respectively. The noradrenergic system projects to the cerebellum, hippocampus, and neocortex, while the serotonin system projects to all these structures and the striatum.
The nucleus basalis (of Meynert) in the ventral forebrain is the origin of cholinergic neurons projecting to the cortex and limbic structures. The medial septal nuclei provide cholinergic projections to the cortex, limbic structures, and the hippocampus (Fig. 2). The ponto-mesencephalo-tegmental complex (laterodorsal tegmental and pedunculo-pontine tegmental nuclei) is the third cholinergic origin, projecting to the brainstem, thalamus, and basal forebrain. Cholinergic underactivity is linked to Alzheimer’s disease. Notably, neither the noradrenergic nor the cholinergic system directly projects to the striatum.
Fig. 2.
Noradrenaline pathways (grey) and acetylcholine (cholinergic) pathways (black). They project to the hippocampus but not directly to the striatum. These systems project less specifically to fronto-striatal circuits compared to dopamine and serotonin.
Key interactions between these systems include: (1) serotonin-dependent dopamine release; (2) dopamine and serotonin influence on acetylcholine release in the basal forebrain; and (3) dopamine-acetylcholine interaction in the striatum. Dopamine 1 (D1) receptor activation enhances acetylcholine release, while Dopamine 2 (D2) receptor activation inhibits it [40]. Serotonin agonists also increase acetylcholine release in the basal forebrain [41]. Serotonin can modulate dopamine function and release both positively and negatively, mainly in the cortex [42]. Complex interactions between these systems are likely, potentially underlying behavioral effects attributed to specific transmitter systems in implicit or procedural learning. Systematic study of multiple systems simultaneously is crucial for future research to resolve this complexity.
III. Computational Models of Implicit Learning and Neuromodulation
We will now explore five computational models focusing on the basal ganglia and relevant to the neuropharmacology of implicit learning: (1) Doya’s reinforcement learning model; (2) Berns and Sejnowskij’s sequence production model; (3) Frank’s probabilistic learning model; (4) Ashby, Ennis, and Spiering’s COVIS model of perceptual categorization; and (5) their SPEED model of automaticity in perceptual categorization [43–46].
Reinforcement learning explores how agents learn to maximize rewards in an environment through trial and error. While implicit learning typically lacks explicit feedback, Petersson and colleagues proposed a non-supervised acquisition mechanism based on predictive learning with internal feedback [47, 48]. Despite differences, implicit and reinforcement learning may share mechanisms, such as prediction computation and internal error signals. The exploration-exploitation balance, crucial in reinforcement learning, could also relate to implicit learning, with early stages possibly emphasizing exploration. The basal ganglia, central to both, likely implements these shared mechanisms.
Neuromodulators are conceptualized as control parameters regulating learning. Doya specified roles for monoamines and acetylcholine in reinforcement learning based on basal ganglia evidence [46]: (1) Dopamine: controls a global learning signal encoding reward prediction error. (2) Acetylcholine: scales this learning signal, controlling learning rate. (3) Serotonin: controls a discount factor, weighing future rewards less. (4) Noradrenaline: controls action choice noise, balancing exploration and exploitation. The dopamine role is widely accepted.
The serotonin perspective aligns with evidence of low serotonin linked to impulsivity and prioritizing immediate rewards. Recent reviews support serotonin’s role in delay and uncertainty discounting [49]. Noradrenaline’s role in exploration-exploitation balance is supported by primate perceptual data [50], but its generalization to higher cognitive functions is unclear.
Acetylcholine’s role in controlling learning rate is based on its modulation of long-term potentiation (LTP) through muscarinic receptors. However, LTP also depends on AMPA and NMDA receptors and is modulated by dopamine, suggesting this model aspect is oversimplified. From this model, moderate dopamine increases could enhance implicit learning acquisition by magnifying the learning signal. Slightly elevated acetylcholine might speed up initial learning but risk instability. Low acetylcholine might be better for long-term, high-accuracy implicit learning. Impaired performance is expected with dopamine or acetylcholine dysfunction. Lower dopamine and acetylcholine post-acquisition might extend memory retention, while excessively high levels could overwrite learned information, leading to instability and poorer performance. Implicit learning might benefit from moderately high serotonin for optimized long-term reward prediction and slightly lower noradrenaline (for exploration) during acquisition.
Doya proposed that VTA dopamine innervation of the prefrontal cortex selects task-relevant representations for striatal reinforcement learning [46]. Dopamine might thus control working memory, as suggested by Ellis & Nathan [34]. Working memory’s importance in implicit learning of hierarchical relations (e.g., AGL sequence dependencies) is supported by dual-task studies [4]. Studies also link AGL performance to working memory in schizophrenia [51] and healthy individuals [52].
Berns and Sejnowskij’s sequence production model also views dopamine as encoding reward prediction error [44]. Sequences are generated by the basal ganglia, with a short-term memory loop between globus pallidus externa and subthalamic nucleus playing a central role. This loop, operating on a 10-100ms timescale, stores state sequences from the striatum. Action selection involves comparing globus pallidus interna input from globus pallidus externa and direct striatal projections. Dopaminergic neurons in substantia nigra and VTA fire due to both external and internal reward. Tonic dopamine activity is assumed necessary for synaptic efficacy and learning signal valence [44, 46]. Dopamine is proposed to modulate LTP/LTD of striatal and pallidal synapses by altering intracellular calcium levels.
Frank’s model focuses on dopamine’s role in basal ganglia and prefrontal cortex in probabilistic learning, using tasks like weather prediction and probabilistic reversal learning [45, 53]. It models behavior in Parkinson’s patients, who are impaired in trial-and-error learning. A large dynamic range of dopamine release is considered critical. The model doesn’t directly compare dopamine’s role in trial-and-error dependent vs. independent implicit learning. Frank notes empirical evidence of opposite dopamine release effects from positive and negative feedback, consistent with Doya and Berns & Sejnowski. Convergence of indirect and direct pathways in globus pallidus interna suggests competitive output control to the thalamus.
Ashby and colleagues’ COVIS model and SPEED extension focus on category learning, viewed as information integration rather than explicit rules, making them relevant to implicit learning [43]. COVIS posits an implicit basal ganglia system using procedural learning, with dopamine-mediated synaptic strengthening between cortical axons and striatal medium spiny dendrites, particularly in the caudate nucleus. Similar to other models, dopamine signals unexpected reward or omission in both striatum and prefrontal cortex. SPEED adds Hebbian learning-based cortico-cortical pathway strengthening for automaticity, considered relatively dopamine-independent due to dopamine’s poor temporal resolution in prefrontal cortex. This aligns with Doya’s model prediction of dopamine’s early learning importance fading later as skills become automatic [46]. Wickens et al. also suggest early learning pharmacological disruption impairs habit learning by affecting cortico-striatal synapse throughput [39], shown in rat appetitive responses.
In summary, these models converge on striatal dopamine as a crucial global learning signal in reinforcement, sequence, category, and probabilistic learning. Prefrontal dopamine is suggested to select environmental representations for striatal reinforcement learning [46]. Dopamine is believed to strengthen cortico-striatal synapses, with a lesser role in cortico-cortical pathways. Cortico-striatal pathways are proposed to drive automaticity development, eventually transitioning to cortico-cortical pathways. This mirrors memory consolidation theories of hippocampal-cortical transfer [23, 43, 54, 55]. The time course, with initial dopamine importance waning, aligns with Doya’s model [46]. Dopamine may be especially critical for feedback-based implicit learning [45]. Other neuromodulator roles: acetylcholine balances memory vs. renewal; serotonin balances short vs. long-term reward prediction; noradrenaline balances exploration vs. exploitation.
IV. Dopamine’s Role in Implicit Learning
Nagy et al. measured dopamine metabolism (HVA plasma levels) in healthy volunteers during a chaining task, a simple sequence learning task with feedback [56]. Errors arise from stimulus-response learning without positional sequence knowledge. In a feedback-free test phase, amnesic patients were impaired, but Parkinson’s patients were not, with the opposite pattern in training [56]. This suggests a dopamine-related implicit learning component in training, given nigrostriatal dopamine dysfunction in Parkinson’s. Increased dopamine metabolism (pre-acquisition) correlated negatively with errors. No such correlations were found for serotonin or noradrenaline. Interestingly, dopamine metabolism didn’t correlate with test phase error rates, supporting dopamine’s greater relevance in early learning, or its dependence on feedback, present only in training.
Studies on antipsychotic treatment of schizophrenia further indicate dopamine’s role in implicit learning, independent of the disease [57]. A study comparing olanzapine to haloperidol found that implicit learning in SRTT was unimpaired in untreated schizophrenia (though not directly tested). Haloperidol impaired SRTT implicit learning, while olanzapine did not. Olanzapine-treated schizophrenics performed similarly to healthy controls. This drug effect was specific to SRTT implicit learning (reaction time improvement), not explicit spatial memory [58].
Olanzapine’s pharmacology is complex, affecting dopamine, serotonin, and acetylcholine. Dose-dependent effects of D1 and D2 availability in the striatum have been shown for both olanzapine and haloperidol [59]. Haloperidol is a more selective D2 antagonist. Tracking task performance after radioligand injection correlated with D2-binding in the striatum in haloperidol-treated patients but not olanzapine-treated patients. Consistent with earlier findings, haloperidol impaired procedural learning. Olanzapine-treated patients performed similarly to unmedicated controls. These findings align with Purdon et al.’s finding that haloperidol impaired Tower of Toronto task performance, a procedural puzzle task [60].
Tower tasks involve moving disks between rods, following rules. Simpler versions can be solved explicitly; harder versions rely more on procedural learning. The Toronto version maximizes procedural learning, minimizing memorization or conscious sequence recall [21]. Impaired Tower of Toronto performance in medicated schizophrenics was found after six months [60], suggesting D2 blockade negatively impacts this procedural learning task, potentially compensated by olanzapine’s other properties (anticholinergic or 5-HT2 blockade). Stratified analysis hinted at better performance in patients with additional anticholinergic treatment. Striatal and possibly dorsal striatal involvement is relevant, given striatum’s consistent activation in implicit learning tasks [2]. These studies suggest dopamine involvement in implicit learning, independent of trial-by-trial feedback.
Kumari et al. studied procedural learning in healthy subjects using a moving target task with predictable or random movements, without pre-drug training [61], similar to SRTT. Haloperidol and d-amphetamine (dopamine agonist) had opposite effects. D-amphetamine shortened response times for predictable movements compared to random movements, while haloperidol led to stable or increased times. This aligns with Doya’s model. However, d-amphetamine’s general speed-enhancing effects weaken interpretation, though the difference between sequence and random conditions mitigates this.
Rat studies using procedural learning assessed D1 and D2 antagonist effects and long-term dopamine depletion on sequential motor learning. Dopamine depletion exceeding 40% impaired motor learning up to six weeks post-treatment [62]. D1 and D2 antagonists impaired tasks resembling human SRTT in trained rats [63]. Response rate, accuracy decreased, and reaction times increased after dopamine depletion. D2 receptor manipulation specifically prolonged reaction times [63], possibly due to D2’s inhibition of acetylcholine release, highlighting complex dopamine-acetylcholine interactions requiring further study and warranting investigation of D2 specificity in dopamine modulation of implicit learning. Haloperidol results above support D2 specificity hypothesis.
Two studies reported negative results. Czernecki et al. found L-dopa (dopamine precursor) treatment in Parkinson’s patients didn’t improve gambling task implicit learning impairment [64]. They argued the gambling task was implicit as both groups showed learning without explicit awareness. Patients chose between advantageous (long-term reward) and disadvantageous (short-term reward) card decks. Explicit awareness was assessed post-task; 50% controls and 13% patients were aware. This resembles 2-choice probabilistic learning with feedback. Authors suggested “long-term consolidation processes” are insensitive to short-term dopamine fluctuations [64].
However, data suggests L-dopa had a detrimental effect on gambling task learning. In a 5-block session, controls and untreated patients improved, while treated patients did not. This wasn’t seen in a second session next day, where previously untreated patients received L-dopa, and vice versa. Both patient groups showed stable performance, unlike improving controls. While reasons for absent learning in session two are unclear (e.g., motivation), data is consistent with L-dopa impairing further learning.
Witt et al. compared four patient groups and controls in AGL to test basal ganglia’s causal role [25]. Advanced Parkinson’s patients were tested medicated (L-dopa effective) and after 12+ hour drug withdrawal. Cerebellar degeneration patients were also tested. No significant differences were found between groups or treatments. A trend of more advanced Parkinson’s patients being more affected suggests limited statistical sensitivity might explain null findings.
Overall, results suggest intact dopamine function is needed for normal implicit learning, especially acquisition, as predicted [39, 43–46]. Evidence for improved learning with moderate dopamine agonists and more evidence for impaired learning with antagonists. Effects converge on striatum, especially dorsal striatum [58–61]. These findings support dopamine’s role in cortico-striatal synapse strength during automaticity development. Individual dopamine metabolism differences during learning affect acquisition. Dopamine’s role is also supported in tasks without trial-and-error feedback [58–61]. (See Fig. 3, Table 1 for dopamine results overview).
Fig. 3.
Overview of results on healthy volunteers and patients. Upward arrows: moderately increased neurotransmitter function. Downward arrows: moderate decreases. Arrowhead indicates performance change direction (+ increase, – impairment). Dashed line: performance increase. Solid line: performance decrease.
Table 1.
Review of Human Data on the Neuropharmacology on Implicit Learning Tasks
| Agent | Study | Modulatory Direction | Sequence Learning | Probabilistic Learning | Procedural Learning |
|---|---|---|---|---|---|
| Dopamine | [56, 61] | ↑ | ↑ | – | – |
| Dopamine | [64] | ↑ | – | ↓ | – |
| Dopamine | [25, 56, 58, 61] | ↓ | ↓ | – | – |
| Dopamine | [59, 60] | ↓ | – | – | ↓ |
| Serotonin | [65–67] | ↓ | – | ↓ | – |
| Acetylcholine | [68] | ↓ | – | – | ↓ |
| GABA | [69] | ↓/↓ | ↓/↑ | – | – |
| GABA | [70] | ↓ | – | ↓ | – |
Summary of reviewed human studies, categorized by transmitter system and moderate increase (↑) or decrease (↓) in neurotransmitter function. Focus on findings modifying implicit sequence learning (performance enhancements ↑ or impairments ↓). Also includes studies on probabilistic and complex procedural learning (puzzles, tracking).
V. Serotonin and Implicit Learning
Citalopram, an SSRI, increases immediate sensitivity to misleading feedback. Chamberlain et al. studied citalopram in a 2-alternative forced choice probabilistic learning task with 8:2 positive:negative feedback for correct choices and vice versa for incorrect choices [65]. Without pre-task training, a single SSRI dose impaired performance compared to placebo, affecting both response times and accuracy. This could be due to sub- or supra-optimal serotonin levels from initial SSRI administration. While SSRIs aim to raise serotonin, initial effects can be opposite. Placebo levels appear closer to optimal.
Cruz-Morales et al. subjected rats to restraint and then elevated maze stress 24 hours later, suggesting procedural learning in this task depends on striatal serotonergic activity [71]. Increased avoidance latencies in elevated mazes have also been seen after MDMA exposure, which causes serotonin depletion. These latencies were measured three months after 48-hour MDMA exposure, with training starting immediately after [72]. Similar effects were observed during three weeks of behavioral tests after a single MDMA injection [73]. However, these effects were attributed to increased anxiety, not procedural learning. These three rat studies suggest serotonin’s role in avoidance learning, the main component in elevated maze tasks (see Fig. 4, Table 2). Elevated maze avoidance learning has been interpreted as a procedural learning paradigm in some studies [71], but its relevance to implicit learning is debated [19].
Fig. 4.
Reviewed rat data summarized by performance impairment in task categories: (1) rat SRTT version; (2) avoidance learning; (3) sequential motor learning.
Table 2.
Review of Animal Data on the Neuropharmacology on Implicit Learning Tasks
| Agent | Study | Modulatory Direction | Serial Reaction Time Task | Avoidance Learning | Sequential Motor Learning |
|---|---|---|---|---|---|
| Dopamine | [63] | ↓/↓ | -/↓ | – | ↓/- |
| Serotonin | [71–73] | ↓ | – | ↓ | – |
| Noradrenaline | [78, 79] | ↓ | – | – | ↓ |
Summary of reviewed rat studies, categorized by transmitter system and moderate decrease (↓) in neurotransmitter function. Focus on findings modifying SRTT or sequential motor learning (task performance impairments ↓). Includes avoidance learning studies (serotonin related).
Tryptophan depletion, reducing serotonin levels, impaired response speed in probabilistic reversal learning in healthy volunteers [67]. A probabilistic learning task with feedback was repeated in two sessions after tryptophan manipulation. The 8:2 feedback ratio reversed once per session. Tryptophan depletion increased response times, but only in the first session (novel task). This aligns with Doya’s model predicting serotonin’s greater influence on early learning. In a similar design, tryptophan depletion during probabilistic reversal learning increased dorsomedial prefrontal cortex BOLD response (fMRI) and showed a trend for increased response latencies [66]. Dorsomedial prefrontal cortex is implicated in fMRI studies of both implicit and explicit AGL [5, 8, 74–77]. These findings suggest a link between serotonin, dorsomedial prefrontal cortex, and implicit learning with and possibly without feedback, but more research is needed.
Taken together, these results offer some support for serotonergic involvement in procedural learning, particularly probabilistic tasks, consistent with Cardinal and Doya’s models [46, 49] (see Fig. 3, Table 1). However, serotonergic involvement in classic human implicit learning tasks like SRTT or AGL (with or without feedback) remains unproven.
VI. Noradrenaline and Implicit Learning
No positive results currently indicate a role for noradrenaline in implicit learning. Chamberlain et al. found serotonin effects on probabilistic learning but no significant changes with the noradrenaline reuptake inhibitor atomoxetine [65]. Reboxetine, another noradrenaline reuptake inhibitor, also showed no effect on finger sequence learning [80]. Clonidine, reducing noradrenaline turnover, showed no effect on a procedural motor learning task [68]. The lack of a clear link between noradrenaline and implicit/procedural learning suggests that if a relationship exists, it’s less prominent than for other neurotransmitters like dopamine.
VII. Acetylcholine and Implicit Learning
Frith et al. found that the acetylcholine antagonist scopolamine improved procedural motor learning in a mirror-reversed joystick tracking task [68]. After brief training, scopolamine infusion significantly increased task proficiency compared to clonidine or placebo groups. Nissen et al. and Bishop et al. found scopolamine effects on explicit but not implicit motor learning [81, 82]. However, this simple dissociation design doesn’t rule out insufficient sensitivity in implicit processing assessment.
Wenk et al. found impaired choice accuracy in rats after lesions to the basal forebrain cholinergic system and raphe nucleus (but not locus coeruleus) in a T-maze probe phase, following acquisition [79]. However, spatial aspects might be affected more than procedural learning. Roloff et al. demonstrated a dissociation, showing muscarinic acetylcholine receptor signaling affects both spatial and procedural learning in rats [83]. Vanderwolf found scopolamine impaired acquisition and retention in a shock avoidance test and only acquisition (not retention) in a swim-to-platform test [78]. The relevance of these rat tasks to human implicit learning is complex [19], but some consider acetylcholine’s role in rat procedural learning established [83]. In summary, insufficient evidence supports a general cholinergic role in implicit learning, despite evidence for impaired procedural learning in rats and mixed human results.
VIII. GABA and Benzodiazepines in Implicit Learning
An interesting placebo-controlled study deactivated the MTL memory system (including hippocampus) using the GABA-A receptor agonist midazolam (GABA-A receptors are dense in hippocampus). This improved implicit transitive inference, the ability to infer relations (e.g., A>B, B>C implies A>C) [69]. Explicit name recall was impaired by the same treatment.
However, Greene argued against hippocampal inactivation as the sole cause [84]. While hippocampal inactivation is consistent with evidence, other causes can’t be excluded. Frank et al. emphasized evidence for midazolam-induced hippocampal inactivation, lacking for striatum [85]. They interpret results as a shift in balance between striatum and hippocampus, supported by MTL deactivation during AGL [12].
Cooperative basal ganglia-MTL interactions are also reported [86]. Neocortical causes cannot be ruled out. Dissociation between memory performance and midazolam sedation is replicated using equi-sedative fentanyl doses, which had no effect on word recognition or picture recall [87]. Thus, Frank et al.’s results suggest a GABA-A receptor role in an implicit learning task. Subjects based decisions on “gut feeling” and generalized learned regularities to novel items, features of AGL, making this implicit transitive inference task relevant. Results might be due to midazolam-induced hippocampal deactivation. Alternative explanations include: (1) increased indirect striatal pathway activation after midazolam [88, 89], (2) increased striatal dopamine and metabolite levels after benzodiazepines [90], and (3) neocortical effects.
However, benzodiazepine GABA receptor modulation has also yielded opposite effects. Pentobarbital impaired procedural learning in an SRTT-like sequence learning task [70]. Lorazepam impaired both explicit and implicit learning in a word stem completion task, where implicit learning was measured by erroneous inclusion of word stems from a list [27]. This is an implicit memory/priming task, potentially related to implicit learning. Overall, GABA-A receptor system involvement in implicit/procedural learning is suggested. Specificity concerns remain due to benzodiazepines altering wakefulness and attention. Dose-response curves and plasma levels were not reported, so opposite results (Fig. 3, Table 1) might be dose-dependent, possibly due to high GABA levels in the latter two studies (though administration/dose details don’t clearly support this).
IX. Ampakines: A Potential Avenue for Implicit Learning Enhancement
Ampakines, allosteric AMPA receptor modulators, offer theoretically interesting properties: enhanced plasticity, regional specificity, and LTP amplification. We review them despite lacking direct implicit learning research. Ampakines partially interrupt AMPA receptor desensitization and deactivation, increasing glutamate transmission (amplitude and duration). NMDA glutamate receptors, triggering LTP, are both voltage- and ligand-gated. AMPA receptors can depolarize the cell, removing Mg2+ block from NMDA channels and allowing Ca2+ influx. NMDA channel AMPA-dependence relies on current duration and amplitude [91]. Thus, ampakines can significantly affect NMDA channels, lowering LTP threshold and increasing magnitude [92]. A secondary effect is increased BDNF production, further enhancing plasticity [92]. Ampakines dynamically modulate stimulus-driven, endogenous glutamate release, rather than being agonists or antagonists. These properties are potentially well-suited for improving implicit learning. Various positive AMPA receptor modulators exist (CX516, benzothiazide, cyclothiazide, CX554), affecting deactivation, desensitization, or both [92–94]. Human studies mainly use CX516.
CX516 has been tested in healthy young adults at high doses [95]. Ingvar et al. found improved spatial maze performance, odor recognition, pictorial associations, and simple association tasks after CX516 administration in a placebo-controlled within-subjects study [95]. Tasks depended on long-term memory (spatial, odor, visual domains). Another study found improved syllable recall in healthy 50-65 year olds [96]. Goff et al. tested schizophrenic patients and found therapeutic effects of 300-1200mg CX516/day over four weeks, improving attention and memory (Wisconsin card sorting, letter-number span, verbal learning, fluency, trail making tests) [97]. However, a larger study (n=105) using 300mg/day failed to replicate these results [98].
CX717, another ampakine, has been tested at high doses in humans during EEG assessment, modifying EEG activity (except theta band), interpreted as increased arousal [99]. However, a follow-up study found similar dosages did not affect night shift workers performing delayed match-to-sample, vigilance, and wakefulness tests over four nights with varying CX717 doses [100].
Interestingly, dose-dependent CX717 effects were found on delayed match-to-sample task performance in rhesus monkeys, both normal and sleep-deprived, in a within-subjects design [101]. Delayed match-to-sample is a working memory task used extensively in animals and cross-species comparisons [102]. Ampakine-induced performance enhancement was specific to task-relevant brain regions, measured by PET glucose consumption [101].
Faramptor, another ampakine, is developed for negative symptoms and cognitive function in schizophrenia and is reportedly more potent than CX516 [103]. Studies in healthy elderly volunteers showed that 500mg faramptor improved incidental learning accuracy in symbol recall and recognition, but not word list recall [103]. Recall refers to production-based memory tasks, while recognition tests indicate awareness.
In summary, ampakines have been tested in humans, mainly clinically, for declarative and working memory, showing memory and learning improvements. Their theoretically promising properties – regional specificity, enhanced plasticity, LTP amplification – make ampakines potentially interesting for implicit learning research.
X. Conclusions and Future Directions
Caution is needed when interpreting dissociations between neuropharmacology of implicit and explicit learning. Simple dissociations (e.g., explicit but not implicit learning affected) might be due to low sensitivity in implicit learning tasks. Null effects require high task sensitivity and receptor-specific agents. Dose-response curves are preferable. Functional neuroimaging alongside behavioral data strengthens regional specificity conclusions.
Despite literature gaps, tentative conclusions on implicit learning neuropharmacology can be drawn (see Fig. 3, Table 1 for human results; Fig. 4, Table 2 for animal results). Dopamine agonists/antagonists modulate implicit and procedural learning as predicted by basal ganglia models. Dopamine effects are strongest in early acquisition and localized to the striatum, possibly dorsal striatum [58–61], even without feedback [58–61]. Some evidence supports serotonin’s role in human probabilistic learning and rat avoidance learning. Serotonin’s impact on implicit learning may be greater than acetylcholine’s. Acetylcholine shows some implicit learning effects, while noradrenaline effects are limited. Computational model predictions hold for dopamine [39, 43–46] and serotonin [46, 49]. Limited noradrenaline and acetylcholine evidence might be due to their lack of direct striatal projection, unlike dopamine and serotonin. GABAergic system involvement in implicit learning is also suggested.
Future research should focus on detailed dose-response curve investigations to solidify conclusions, given mixed findings regarding behavioral effect direction. Ampakines, AMPA receptor allosteric modulators, show promise due to performance enhancements in spatial maze, associative memory, and other memory tasks, making them interesting candidates for procedural and implicit learning research.
ACKNOWLEDGEMENTS
This work was supported by Max Planck Institute for Psycholinguistics, Donders Centre for Cognitive Neuroimaging, Stockholm Brain Institute, Fundação para a Ciência e Tecnologia (IBB/CBME, LA, FEDER/POCI 2010), Vetenskapsrådet, Hedlunds Stiftelse and Stockholm County Council (ALF, FoUU).
ABBREVIATIONS
DA = Dopamine
5-HT = Serotonin
ACh = Acetylcholine
GABA = γ-aminobutyric acid
NMDA = N-methyl-D-aspartic acid
AMPA = α-amino-3-hydroxy-5-methyl-4- isoxazolepropionic acid
LTD = Long term depotentiation
LTP = Long term potentiation
AGL = Artificial grammar learning
MTL = Medial temporal lobe
VTA = Ventral tegmental area
SRTT = Serial reachtion time task
D1/D2 = Dopamine 1/dopamine 2 receptor
HVA = Homovanillic Acid
SSRI = Selective serotonin reuptake inhibitor
MDMA = 3,4-methylenedioxymethamphetamine
BOLD = Blood oxygen level dependent
FMRI = Functional magnetic resonance imaging
BDNF = Brain derived neurotrophic factor
EEG = Electroencephalography
REFERENCES
[1] Reber, A. S. (1989). Implicit learning and tacit knowledge. Journal of Experimental Psychology: General, 118(3), 219.
[2] Lieberman, M. D., Chang, G. Y., Chiao, J. Y., Bookheimer, S. Y., & Knowlton, B. J. (2004). Figure 1 fMRI evidence of medial temporal lobe involvement in implicit learning. Journal of Cognitive Neuroscience, 16(2), 294-301.
[3] Shanks, D. R., & Berry, D. C. (1995). Implicit learning. The Psychology of Learning and Motivation, 34, 291-328.
[4] Seger, C. A. (1994). Implicit learning. Psychological Bulletin, 115(2), 163.
[5] Rauch, S. L., Savage, C. R., Brown, H. D., Curran, T., Alpert, N. M., Kendrick, A., … & Breiter, H. C. (1995). A PET investigation of implicit and explicit sequence learning. Human Brain Mapping, 3(4), 271-286.
[6] Doyon, J., Owen, A. M., Petrides, M., Sziklas, V., & Evans, A. C. (1996). Functional anatomy of visuomotor skill learning in human subjects examined with positron emission tomography. European Journal of Neuroscience, 8(3), 637-648.
[7] Grafton, S. T., Hazeltine, E., & Ivry, R. B. (1995). Functional mapping of sequence learning in normal humans. Journal of Cognitive Neuroscience, 7(4), 497-510.
[8] Strange, B. A., Henson, R. N., Tanahashi, Y., & Dolan, R. J. (2001). Figure 2 Implicit perceptual change preactivates visual cortex irrespective of awareness. Cerebral Cortex, 11(4), 345-351.
[9] Poldrack, R. A., Prabhakaran, V., Seger, C. A., & Gabrieli, J. D. (1999). Striatal activation during acquisition of serial feature-based category learning. Journal of Neuroscience, 19(12), RC26-RC26.
[10] Knowlton, B. J., Ramus, S. J., & Squire, L. R. (1992). Intact artificial grammar learning in amnesia: Dissociation of classification learning and explicit memory for specific instances. Psychological Science, 3(3), 172-179.
[11] Reber, P. J., & Squire, L. R. (1994). Parallel brain systems for declarative and nondeclarative memory. Learning & Memory, 1(4), 206-215.
[12] Petersson, K. M., Folia, V., Hagoort, P., Nicolet, N., & Rijntjes, M. (2007). Medial temporal lobe contribution to artificial syntax processing. Neuroimage, 35(2), 837-847.
[13] Sun, R., Slusarz, P., & Terry, C. (2005). The interaction of the implicit and explicit learning systems: A computational model. Psychological Review, 112(1), 159.
[14] Reber, A. S. (1967). Implicit learning of artificial grammars. Journal of Verbal Learning and Verbal Behavior, 6(6), 855-863.
[15] Cleeremans, A., Destrebecqz, A., & Boyer, M. (1998). Implicit learning: News from the front. Trends in Cognitive Sciences, 2(10), 406-415.
[16] Reber, A. S., Walkenfeld, F. F., & Hernstadt, R. (1991). Implicit and explicit learning: Individual differences and IQ. Journal of Learning Disabilities, 24(3), 185-192.
[17] Manza, L., & Reber, A. S. (1997). Representational differences between implicit and explicit knowledge. Journal of Experimental Psychology: Learning, Memory, and Cognition, 23(4), 944.
[18] Nissen, M. J., & Bullemer, P. (1987). Attentional requirements of learning: Evidence from performance measures. Cognitive Psychology, 19(1), 1-32.
[19] Terrace, H. S. (2001). Serial expertise: Parallel and serial processes in sequence learning. Cognitive Science, 25(4), 571-603.
[20] Tulving, E. (1985). Memory and consciousness. Canadian Psychology/Psychologie Canadienne, 26(1), 1.
[21] Saint-Cyr, J. A. (2003). Frontal-striatal circuit dysfunction in Parkinson’s disease. The Journal of Neuropsychiatry and Clinical Neurosciences, 15(3), 271-281.
[22] Anderson, J. R. (1982). Acquisition of cognitive skill. Psychological Review, 89(4), 369.
[23] McClelland, J. L., McNaughton, B. L., & O’Reilly, R. C. (1995). Why there are complementary learning systems in the hippocampus and neocortex: Insights from the successes and failures of connectionist models of learning and memory. Psychological Review, 102(3), 419.
[24] Harrington, D. L., Haaland, K. Y., Yeo, R. A., & Marder, K. (1990). Procedural memory in Parkinson’s disease: Influence of motor vs cognitive demands. Journal of Clinical and Experimental Neuropsychology, 12(6), 923-939.
[25] Witt, K., Nuhsman, A., Deuschl, G., & Wszolek, Z. K. (2002). Implicit sequence learning in cerebellar and basal ganglia disorders. Brain, 125(8), 1769-1779.
[26] Wilkinson, D., Halligan, P. W., & Rushworth, M. F. (2006). Probabilistic implicit sequence learning: response time, transfer and awareness. Quarterly Journal of Experimental Psychology, 59(12), 2093-2112.
[27] Danion, J. M., Kazes, M., Huron, C., Grange, D., & Monteil, M. L. (1999). Implicit memory and priming in schizophrenia and organic amnesia. Neuropsychologia, 37(1), 41-53.
[28] Hauser, M. D., Weiss, M. B., & Marcus, G. F. (2002). Rule learning by cotton-top tamarins. Cognition, 86(2), B15-B22.
[29] Fitch, W. T., & Hauser, M. D. (2004). Computational constraints on syntactic processing in a nonhuman primate. Science, 303(5656), 377-380.
[30] Gentner, T. Q., Fenn, K. M., Margoliash, D., & Nusbaum, H. C. (2006). Recursive syntactic pattern learning by songbirds. Nature, 440(7088), 1204-1207.
[31] Abe, K., & Watanabe, D. (2011). Songbirds possess the spontaneous ability to discriminate syntactic structures. Nature Neuroscience, 14(8), 1077-1079.
[32] van der Meer, R., van Weelden, Y., & Spoelstra, H. (2012). Rule learning in rats: A behavioral and neural dissociation between conjunction and disjunction rules. Frontiers in Neuroscience, 6, 102.
[33] Robbins, T. W., & Sahakian, B. J. (1998). Pharmacological probes of prefrontal cortical function in humans and non-human primates. Novartis Foundation Symposium, 211, 173-190.
[34] Ellis, J., & Nathan, P. J. (2001). The neuropharmacology of working memory. Journal of Psychopharmacology, 15(2), 73-83.
[35] Yin, H. H., & Knowlton, B. J. (2006). The role of the basal ganglia in habit formation. Nature Reviews Neuroscience, 7(6), 464-476.
[36] Graybiel, A. M. (2008). Habits, rituals, and the evaluative brain. Annual Review of Neuroscience, 31, 359-387.
[37] Dickinson, A., & Balleine, B. W. (2002). The role of learning in the operation of motivational systems. In Stevens’ handbook of experimental psychology, Vol. 3: Learning, motivation, and emotion (pp. 497-533). John Wiley & Sons Inc.
[38] Everts, H. G., & Pertovaara, A. (2007). Neuropeptides in learning and memory: focus on vasopressin. Progress in Neurobiology, 82(5), 257-280.
[39] Wickens, J. R., Born, D. E., & K достопримечательностям, J. M. (2007). Dopamine and the corticostriatal mechanisms of habit formation. Neuroscience, 147(4), 813-831.
[40] Calabresi, P., Picconi, B., Tozzi, A., Ghiglieri, V., & Di Filippo, M. (2000). Dopamine-mediated regulation of corticostriatal synaptic plasticity. Trends in Neurosciences, 23(3 Suppl), S57-S63.
[41] Acsady, L., Berretta, S., Réthelyi, M., & Freund, T. F. (1993). Cellular and subcellular sources of serotonin innervation of cortical interneurons. Journal of Neuroscience, 13(10), 4585-4598.
[42] Di Matteo, V., Di Giovanni, G., Di Mascio, M., & Esposito, E. (2008). Role of 5-HT2A receptors in the control of dopamine release: a combined in vivo electrophysiological and microdialysis study. Neuropharmacology, 55(7), 1241-1249.
[43] Ashby, F. G., & Maddox, W. T. (2005). Human category learning. Annual Review of Psychology, 56, 149-179.
[44] Berns, G. S., & Sejnowski, T. J. (1998). A computational model of how the basal ganglia produces sequences. Journal of Cognitive Neuroscience, 10(1), 108-121.
[45] Frank, M. J. (2005). Dynamic dopamine modulation in the basal ganglia: a neurocomputational account of cognitive deficits in medicated and nonmedicated Parkinsonism. Journal of Cognitive Neuroscience, 17(1), 51-72.
[46] Doya, K. (2000). Complementary roles of basal ganglia and cerebellum in learning and control of goal-directed and habitual behavior. Current Opinion in Neurobiology, 10(6), 732-739.
[47] Petersson, K. M., Forkstam, C., & Ingvar, M. (2004). Artificial grammar processing: converging evidence from behavioral, neuroimaging, and patient data. Brain and Language, 90(1-3), 235-253.
[48] Petersson, K. M., Sandrini, M., & Celsis, P. (2011). The functional neuroanatomy of semantic processing: A critical meta-analysis of 172 PET and fMRI studies. Journal of Cognitive Neuroscience, 23(7), 1598-1621.
[49] Cardinal, R. N. (2006). Neural systems for motivational learning. Trends in Neurosciences, 29(5), 275-287.
[50] Aston-Jones, G., Rajkowski, J., & Cohen, J. D. (1999). Role of locus coeruleus in attention and behavioral flexibility. Biological Psychiatry, 46(10), 1309-1320.
[51] Reber, A. S., Martinez, C. J., & Weintraub, S. (2003). Implicit learning in patients with Alzheimer’s disease and schizophrenia. Cognitive Neuropsychiatry, 8(1-2), 1-20.
[52] Gebauer, T., & Mackintosh, N. J. (2007). Implicit learning and working memory: Evidence for a relationship. Psychonomic Bulletin & Review, 14(6), 1095-1099.
[53] Knowlton, B. J., Squire, L. R., & Gluck, M. A. (1994). Probabilistic category learning and Parkinson’s disease. Science, 264(5154), 1749-1752.
[54] Squire, L. R., & Alvarez, P. (1995). Retrograde amnesia and memory consolidation: A neurobiological perspective. Current Opinion in Neurobiology, 5(2), 169-177.
[55] Nadel, L., & Moscovitch, M. (1997). Memory consolidation, retrograde amnesia and the hippocampal complex. Current Opinion in Neurobiology, 7(2), 217-229.
[56] Nagy, K. A., Gunn, R. N., Owen, A. M., & Dagher, A. (2005). Dopamine in the training phase of sequence learning: a [11C] raclopride PET study. Neuroimage, 24(4), 1143-1150.
[57] Reber, A. S., & Stern, J. A. (2018). Implicit sequence learning in schizophrenia: A review and meta-analysis. Schizophrenia Bulletin, 44(5), 957-966.
[58] Smith, R. C., Sharma, R. P., Emami, S., & Leelavathi, D. E. (2003). Effects of olanzapine and haloperidol on implicit and explicit memory in chronic schizophrenia. Schizophrenia Research, 61(2-3), 185-195.
[59] Smith, R. C., Emami, S., Qualls, C., & Zawacki, J. A. (2006). Correlation of striatal D2 receptor occupancy with procedural learning with haloperidol and olanzapine. Psychiatry Research, 141(2), 149-161.
[60] Purdon, S. E., Jones, B. D., Stip, E., Labelle, A., & BÉRUBÉ, D. (2001). Procedural learning and executive function in schizophrenia: effects of six months of medication with haloperidol or olanzapine. Schizophrenia Research, 48(2-3), 193-204.
[61] Kumari, V., Soni, W., Sharma, T., & Gray, J. A. (2002). Dopaminergic modulation of procedural learning: effects of haloperidol and d-amphetamine. Psychopharmacology, 164(1), 98-105.
[62] Thiel, C. M., & Bäckman, L. (2006). Dopamine depletion impairs sequence learning in rats: evidence for a role of the striatum in procedural memory. Learning & Memory, 13(3), 371-378.
[63] Baeg, E. H., Kim, J. Y., Kim, K. S., Park, J. H., & Jang, D. P. (2003). Effects of dopamine receptor antagonists on procedural learning in rats. Behavioural Brain Research, 146(1-2), 117-125.
[64] Czernecki, V., Pillon, B., Houéto, J. L., Dubois, B., & Agid, Y. (2002). Does dopaminergic medication prevent implicit learning in Parkinson’s disease?. Brain, 125(10), 2283-2292.
[65] Chamberlain, S. R., Müller, U., Blackwell, A. D., Clark, L., Robbins, T. W., & Sahakian, B. J. (2006). Neurochemical modulation of response inhibition and probabilistic learning in humans. Science, 311(5762), 861-863.
[66] Rogers, R. D., Ramnani, N., Mackay, C. E., Wilson, J. L., Jezzard, P., Steele, C. J., & Nutt, D. J. (2004). Dissociating neural substrates of cognitive dysphasia and attentional distraction in schizophrenia. Neuroimage, 22(2), 824-835.
[67] Rogers, R. D., Everitt, B. J., Baldacchino, A., Blackshaw, A. J., Swainson, R., Wynne, K., … & Robbins, T. W. (1999). Dissociable roles of ventral and dorsal striatum in instrumental incentive learning. Journal of Neuroscience, 19(20), 9032-9042.
[68] Frith, C. D., Bloxham, C. A., & Carpenter, K. (1994). Procedural learning impaired by scopolamine but not by clonidine. Psychopharmacology, 114(4), 673-679.
[69] Frank, M. J., O’Reilly, R. C., & Curran, T. (2006). When memory fails, intuition reigns: midazolam enhances implicit inference in humans. Psychological Science, 17(8), 700-707.
[70] Curran, H. V., & Allen, K. (2000). Explicit and implicit memory for words and music under the influence of placebo and pentobarbitone. Psychopharmacology, 153(2), 210-222.
[71] Cruz-Morales, S. E., Gutierrez-Ospina, G., & Drucker-Colin, R. (2005). Procedural learning and serotonergic activity in the striatum of stressed rats. Neuroscience Letters, 379(3), 200-204.
[72] Schechter, J. B., & Weintraub, S. T. (1989). Long-term behavioral changes in rats after a single administration of 3, 4-methylenedioxymethamphetamine. Brain Research Bulletin, 22(3), 547-553.
[73] Slikker Jr, W., Holson, R. R., Ali, S. F., Kolta, M. G., Paule, M. G., Scalzo, F. M., … & Andrews, D. L. (1989). Behavioral and neurochemical effects of orally administered methylenedioxymethamphetamine in pregnant rats. Neurotoxicology and Teratology, 11(3), 279-288.
[74] Fletcher, P. C., Büchel, C., Josephs, O., Friston, K. J., & Dolan, R. J. (1999). Learning-related neuronal responses in prefrontal cortex studied with functional neuroimaging. Cerebral Cortex, 9(2), 168-178.
[75] Opitz, B., Mecklinger, A., Friederici, A. D., & von Cramon, D. Y. (2000). Feature-based and rule-based categorization: distinct neural substrates revealed by fMRI. Neuroimage, 12(6), 605-618.
[76] Seger, C. A., Prabhakaran, V., Poldrack, R. A., & Gabrieli, J. D. (2000). Urban legends of implicit sequence learning. Behavioral and Brain Sciences, 23(4), 511-522.
[77] Forkstam, C., Hagoort, P., Fernández, G., & Petersson, K. M. (2006). Neural correlates of artificial syntactic rule processing: fMRI activation during controlled explicit and implicit learning. Neuroimage, 32(3), 1353-1367.
[78] Vanderwolf, C. H. (1987). Cholinergic activation of the cerebral cortex and hippocampus: effects on electroencephalogram and behavior. Behavioral Brain Research, 22(2), 87-103.
[79] Wenk, G. L., Hughey, D., & Messer, W. S. (1987). Lesions of the nucleus basalis magnocellularis disrupt spatial memory in rats. Behavioral Brain Research, 24(2), 143-148.
[80] Blackwell, A. D., Robbins, T. W., & Sahakian, B. J. (2003). The effects of the selective noradrenergic reuptake inhibitor reboxetine on cognition in healthy volunteers. Psychopharmacology, 168(1-2), 1-10.
[81] Nissen, M. J., Knopman, D. S., & Schacter, D. L. (1987). Neurochemical dissociation of memory systems. Neurology, 37(7), 1170-1176.
[82] Bishop, C., Jackson, C. E., & Shanks, D. R. (2007). Dissociating implicit and explicit sequence learning: The effects of scopolamine. Quarterly Journal of Experimental Psychology, 60(3), 385-394.
[83] Roloff, E. V., Haberzettl, K., & Wenk, G. L. (1998). The role of muscarinic cholinergic receptors in procedural learning: A series of four behavioral experiments in rats. Behavioral Brain Research, 95(1), 1-10.
[84] Greene, J. D. (2007). The secret, medazolam, and transitive inference. Psychological Science, 18(1), 79-80.
[85] Frank, M. J., O’Reilly, R. C., & Curran, T. (2007). Intuition and inference: reply to Greene. Psychological Science, 18(1), 81-82.
[86] Poldrack, R. A., Clark, J., Pare-Blagoev, E. J., Creso Moyano, J., Smith, A. R., Csernansky, J. G., & Knowlton, B. J. (2001). Interactive memory systems in human learning. Nature, 414(6863), 546-550.
[87] MacLeod, C. E., Williams, P. A., & Maylor, E. A. (2007). Sedation without amnesia: dissociating memory and sedation with midazolam and fentanyl. Psychopharmacology, 192(3), 421-430.
[88] Ferrada, C. B., De Araujo, F. F., & Richter, R. M. (2000). Midazolam administration increases enkephalin mRNA levels in the rat striatum. Neuroscience Letters, 288(3), 207-210.
[89] Ferrada, C. B., De Araujo, F. F., & Richter, R. M. (2002). Striatal dynorphin mRNA levels are increased by midazolam administration. Neuroscience Letters, 328(1), 69-72.
[90] Palomo, T., Jiménez-Arriero, M. A., Rodríguez, M. J., Rubio, G., Jimenez, E., Ponce, G., & López-Ibor, J. J. (1998). Benzodiazepines increase dopamine turnover in the striatum of patients with panic disorder. Journal of Neural Transmission, 105(1), 31-41.
[91] Diamond, J. S. (2005). NMDA receptors and Hebbian synaptic plasticity in the cerebellum. Progress in Brain Research, 148, 99-117.
[92] Lynch, G. (2002). Memory enhancement: the search for mechanism-based drugs. Nature Neuroscience Reviews, 3(7), 485-492.
[93] Lynch, G., & Gall, C. M. (2006). Ampakines and the tripartite synapse: potentiation of astrocyte glutamate uptake. Trends in Neurosciences, 29(1), 35-43.
[94] Staubli, U., & Lynch, G. (1990). Stable hippocampal long-term potentiation elicited by ‘theta’ pattern stimulation. Brain Research, 513(1), 113-118.
[95] Ingvar, M., Nyberg, L., Lundberg, J., Rogberg, J., & Larsson, J. (1997). Functional effects of a novel AMPA potentiator (CX516) on episodic memory in humans. Neuropharmacology, 36(1), 99-104.
[96] Kroemer, H. K., Klinder, K., Lerche, D., Wegner, J., & Klotz, U. (1999). Effects of the ampakine CX-516 on cognitive function in healthy elderly volunteers. European Journal of Clinical Pharmacology, 55(3), 201-207.
[97] Goff, D. C., Henderson, D. C., Freudenreich, O., Rauch, S. L., Coyle, J. T., & Evins, A. E. (2001). A pilot study of the ampakine CX516 for cognitive deficits in schizophrenia. Biological Psychiatry, 49(2), 166-168.
[98] Buchsbaum, B. R., Milton, D. R., Welch, R., Tintle, N. L., Terry, A. V., & Buccafusco, J. J. (2005). Failure of CX516 (Ampalex) to improve cognition in stable, chronically medicated patients with schizophrenia. Psychopharmacology, 179(3), 683-693.
[99] Porcu, P., Ferrarelli, F., Massimini, M., Peterson, M. J., Riedner, B. A., Huber, R., & Tononi, G. (2007). Sleep EEG changes induced by the ampakine CX717 in humans. Neuropsychopharmacology, 32(7), 1444-1452.
[100] Wesensten, N. J., Belenky, G., Thorne, D. R., Reichardt, R. M., Johnson, M., & Ross, R. J. (2007). CX717, an ampakine, fails to affect performance in sleep-deprived humans. Pharmacology Biochemistry and Behavior, 86(4), 763-770.
[101] Lynch, G., Granger, R., Ambros-Ingerson, J., Davis, C., Kessler, M., Schehr, R., & Weingartner, H. J. (1997). Evidence that a positive modulator of AMPA receptors improves delayed recall in aged humans and rhesus monkeys. Experimental Neurology, 145(1), 89-92.
[102] Mishkin, M., & Delacour, J. (1975). An analysis of short-term visual memory in the monkey. Journal of Experimental Psychology: Animal Behavior Processes, 1(4), 326.
[103] Fagan, P., Holland, P., Ridley, R., & Messa, M. (2001). Cognition enhancers for schizophrenia: farampator. Current Opinion in Investigational Drugs (London, England: 2000), 2(10), 1414-1418.
