Trace Conditioning vs. Delay Conditioning

The ability of an organism to adaptively anticipate biologically significant events based on environmental precursors represents one of the most fundamental evolutionary specializations in the animal kingdom. At the core of this anticipatory capacity lies Pavlovian, or classical, conditioning—a behavioral paradigm wherein an initially neutral conditioned stimulus (CS) comes to elicit a conditioned response (CR) following systematic pairings with an intrinsically reinforcing unconditioned stimulus (US). While early twentieth-century behaviorism conceptualized this learning as an automatic consequence of sensory contiguity, modern behavioral neuroscience has demonstrated that classical conditioning is not a monolithic physiological entity. Instead, the computational, cognitive, and neuroanatomical demands of associative learning are dictated by the precise temporal geometry of the stimulus presentation.

Among the various temporal arrangements utilized in empirical investigations, the dichotomy between delay conditioning and trace conditioning provides a classic double dissociation in cognitive neurobiology. In delay conditioning, the CS precedes and overlaps temporally with the US, terminating either concurrently with or after the reinforcer’s onset. In trace conditioning, an explicit, stimulus-free temporal gap—termed the trace interval—intervenes between the offset of the CS and the onset of the US. While this operational variation might appear to be a trivial shift in stimulus timing, its behavioral and neurobiological ramifications are profound. A temporal disparity measured in mere hundreds of milliseconds fundamentally alters the underlying neural circuitry, shifting the burden of associative plasticity from autonomous subcortical and cerebellar loops to complex, distributed networks involving the hippocampus, the medial prefrontal cortex, and associated forebrain structures.

This comprehensive treatise investigates the theoretical, neurobiological, and clinical dimensions that demarcate trace conditioning from delay conditioning. By exploring their temporal architectures, cellular and synaptic substrates, requirements for declarative awareness, and differential susceptibilities to neuropathology, this analysis illustrates how these two paradigms serve as premier experimental tools. Through decades of rigorous behavioral assays, optogenetic interventions, electrophysiological recordings, and clinical patient studies, the contrast between trace and delay conditioning continues to illuminate the mechanisms through which the central nervous system bridges time, integrates sensory information, and transforms fleeting environmental cues into enduring behavioral adaptations.

1. Introduction to Temporal Paradigms in Classical Conditioning

1.1 Defining Classical Conditioning and the Critical Variable of Timing

Classical conditioning, originally systematized by Ivan Petrovich Pavlov during his pioneering investigations into digestive secretions, represents a foundational mechanism through which organisms extract predictive causal relationships from their sensory environment. In canonical Pavlovian paradigms, an organism is repeatedly exposed to a neutral conditioned stimulus—such as an auditory pure tone or a visual illumination—paired in a predictive temporal sequence with an unconditioned stimulus that reflexively evokes an unconditioned response (UR), such as a corneal airpuff evoking an eyeblink or an intraoral delivery of sucrose evoking salivation. Across successive pairings, the organism demonstrates associative learning by executing a conditioned response (CR) prior to the onset of the US, revealing that the CS has acquired acquired behavioral salience and predictive value.

Although initial formulations within associative learning theory posited that simple temporal contiguity—the mere co-occurrence of two physical events in close temporal proximity—was both necessary and sufficient for associative bond formation, modern neurobiology has revealed that timing is not merely a quantitative facilitator of learning, but rather an absolute qualitative determinant. The interstimulus interval (ISI), defined specifically as the temporal duration between the onset of the conditioned stimulus and the onset of the unconditioned stimulus, profoundly dictates the magnitude, rate of acquisition, and morphological precision of the conditioned response. If the ISI is excessively brief (for instance, simultaneous presentation where ISI equals zero), associative learning is typically impoverished or non-existent, as the CS provides zero forward predictive utility to the organism.

Conversely, as the ISI is elongated, the cognitive and neurocomputational resources required to register the predictive association undergo a drastic transformation. The chronological relationship between the stimuli dictates whether the sensory processing systems can rely on the continuous co-activation of peripheral and subcortical sensory channels or whether higher-order mnemonic systems must be recruited to bridge an absolute void of physical stimulation. Consequently, the temporal dimension functions as a biological gatekeeper, determining which neural circuits are engaged to encode, consolidate, and retrieve the predictive associative engram.

1.2 Conceptualizing Trace vs. Delay Conditioning

The operational taxonomy of forward Pavlovian conditioning is structured primarily around two archetypal paradigms: delay conditioning and trace conditioning. The operational definition of delay conditioning rests upon the temporal continuity and direct physical overlap between the conditioned stimulus and the unconditioned stimulus. In this arrangement, the CS is presented first, remains active throughout the duration of the interstimulus interval, and either coterminates precisely with the onset or termination of the US, or persists alongside the US for a brief period. Because the CS does not cease before the US is delivered, there is zero temporal gap separating the physical presence of the two stimuli. The sensory representation of the CS is continuously present within primary sensory reception systems when the unconditioned stimulus arrives.

In striking operational contrast, trace conditioning is defined by the insertion of an explicit, stimulus-free temporal gap between the offset of the conditioned stimulus and the onset of the unconditioned stimulus. The CS is initiated, persists for an assigned duration, and is then completely terminated. The organism then traverses a silent, physical void—the trace interval—before the unconditioned stimulus is delivered. In this paradigm, because the CS is physically absent at the moment the US occurs, the central nervous system cannot rely on concurrent peripheral sensory transduction to link the two events. Instead, the organism must generate, maintain, and access an internal physiological or mnemonic representation—traditionally designated as a “stimulus trace”—across the intervening temporal hiatus to successfully bridge the associative gap.

Significant theoretical and empirical confusion frequently arises regarding the distinction between the overall interstimulus interval and the trace interval. The ISI encompasses the entire temporal distance from CS onset to US onset, whereas the trace interval refers exclusively to the duration from CS offset to US onset. Therefore, two experiments may feature an identical ISI (for example, 750 milliseconds), yet represent entirely different psychological operations: a delay paradigm utilizing a continuous 750-millisecond tone that coterminates with an airpuff, versus a trace paradigm utilizing a 250-millisecond tone followed by an empty 500-millisecond trace interval prior to airpuff delivery. It is the existence of this physical discontinuity, rather than the raw metric of the ISI alone, that imposes fundamentally different neurobiological requirements upon the organism.

1.3 Significance in Behavioral Neuroscience and Cognitive Psychology

The operational divergence between delay and trace conditioning has elevated these paradigms into indispensable methodological assays across cognitive psychology and behavioral neuroscience. Foremost among their experimental utilities is their demonstrated capacity to functionally dissociate subcortical and cerebellar learning systems from higher-order forebrain and telencephalic memory networks. The discovery that standard delay conditioning can be successfully mastered by decerebrate or hippocampal-lesioned organisms, whereas trace conditioning is devastated by identical forebrain disruptions, provided early, incontrovertible neurobiological proof that distinct anatomical circuits mediate phenomenologically similar behaviors depending exclusively on temporal architecture.

Furthermore, the comparison between trace and delay paradigms has served as an empirical cornerstone in exploring the neurobiology of consciousness, attention, and declarative awareness. In human clinical and psychophysiological investigations, delay conditioning typically proceeds automatically, implicitly, and without the prerequisite of conscious contingency awareness, persisting even under conditions of high cognitive distraction or profound amnesia. In contrast, trace conditioning exhibits a stringent dependency on explicit, declarative awareness of the CS–US relationship, demanding attentional allocation and active working memory buffering to span the temporal divide.

In translational medicine and clinical neurobiology, these paradigms offer critical, non-invasive diagnostic probes for assessing neurodegenerative and psychiatric conditions. Because the anatomical circuits mediating both paradigms are mapped with cellular and synaptic precision—spanning the cerebellum, brainstem nuclei, hippocampus, entorhinal cortex, and medial prefrontal cortex—selective impairments in trace versus delay conditioning function as behavioral biomarkers. As a consequence, these temporal variants are routinely deployed to evaluate the progression of early-stage Alzheimer’s disease, schizophrenia, autism spectrum disorders, and age-related cognitive decline, offering unmatched mechanistic resolution into the integrity of both subcortical reflex circuits and complex forebrain memory systems.

2. Comprehensive Overview of Trace Conditioning

2.1 Temporal Architecture and Paradigmatic Structure

The temporal architecture of trace conditioning is defined by its non-overlapping, sequential arrangement of stimuli. The structural chronology unfolds in three distinct, highly regulated epochs: the conditioned stimulus presentation, the stimulus-free trace interval, and the unconditioned stimulus presentation. The trial initiates with the activation of the CS, which is maintained for a predetermined duration (typically ranging from 50 to 500 milliseconds in standard rodent and human eyeblink protocols). At the conclusion of this duration, the CS is turned off completely. The subsequent epoch is the trace interval—an operational void characterized by the complete absence of nominal experimental stimuli, spanning anywhere from a few hundred milliseconds to multiple seconds or, in visceral paradigms such as conditioned taste aversion, several hours.

The parametric properties of the trace interval exert profound, deterministic influences over the rate of associative acquisition and the ultimate behavioral topography of the conditioned response. Experimental data demonstrate an inverse functional relationship between trace interval duration and learning efficacy: as the stimulus-free gap is progressively lengthened, the number of trials required to achieve behavioral criterion increases exponentially, until a threshold is crossed beyond which standard conditioning completely fails to manifest. This decay of associative efficacy reflects the biological constraints of maintaining persistent neural activity within cortical-hippocampal ensembles without external sensory reinforcement.

Crucially, despite the physical absence of the CS during the trace interval, an adeptly conditioned organism does not execute its conditioned response immediately upon CS offset. Instead, trace conditioning displays temporal precision: the organism withholds the execution of the motor or autonomic CR until the precise moment that coincides with the anticipated arrival of the US. This reveals that the trace paradigm requires not merely the preservation of stimulus identity across an empty gap, but also an internal timing mechanism that measures the exact metric length of the trace interval, ensuring adaptive behavioral deployment.

2.2 Neurobiological Substrates and Circuitry

The neurobiological execution of trace conditioning requires an elaborate, multiregional network wherein higher-order forebrain structures dynamically interface with subcortical and cerebellar execution pathways. The fundamental structure required for the acquisition of trace associations across a broad spectrum of sensory modalities is the medial temporal lobe, with an obligate reliance on the hippocampal formation. Bilateral aspiration, neurotoxic, or optogenetic disruptions of the dorsal and ventral hippocampus consistently abolish the acquisition of trace conditioning while leaving delay conditioning entirely undisturbed.

Within the hippocampal circuit, the CA1 and CA3 pyramidal subfields are central to spanning temporal discontinuities. Electrophysiological recordings in awake, behaving animals reveal that specific subsets of CA1 pyramidal neurons exhibit sustained, elevated firing rates throughout the stimulus-free trace interval. This persistent firing—often referred to as “time-cell” activity—is not a passive sensory reverberation, but an active, metabolically demanding physiological bridge. The maintenance of this persistent activity is critically mediated by intrinsic cellular properties, including non-specific cation currents (such as $I_{\text{CAN}}$) activated by cholinergic input from the medial septum, alongside intact N-methyl-D-aspartate (NMDA) receptor-mediated neurotransmission and local microcircuit disinhibition mediated by GABAergic interneuron populations.

Concurrently, the medial prefrontal cortex (mPFC)—including the prelimbic, infralimbic, and anterior cingulate cortices in rodents, and the dorsolateral and ventromedial prefrontal cortices in humans—serves as an indispensable cortical partner to the hippocampus. The mPFC provides the working memory networks and executive top-down control necessary to sustain the internal representation of the CS against decay and interference. Electrophysiological data confirm strong, phase-locked theta-band (4–8 Hz) coherence between the hippocampus and the prefrontal cortex during the trace interval, indicating that high-fidelity oscillatory synchronization underpins the transfer and maintenance of mnemonic representations.

Crucially, while the forebrain network is responsible for encoding and holding the trace representation, it does not act as the final motor execution site. Instead, the hippocampus and prefrontal cortex project directly and indirectly—via the basal ganglia, entorhinal cortex, and deep pontine nuclei—downward to the cerebellum. The cerebellar cortex and deep cerebellar nuclei (predominantly the anterior interpositus nucleus) remain the indispensable motor execution centers that calculate the exact temporal topography of the CR. Thus, trace conditioning represents an advanced, hierarchical system wherein telencephalic networks bridge the temporal void and feed an organized timing signal into the cerebellar apparatus for motor output.

2.3 Cognitive and Informational Requirements

The introduction of a temporal discontinuity forces the central nervous system to shift from simple sensory transduction to complex cognitive representation. From an informational processing perspective, trace conditioning cannot proceed through direct, mechanical associative bonding between concurrently active sensory afferents. Instead, the internal representation of the CS must be converted into an active working memory buffer that endures beyond the physical duration of the sensory event. The organism must essentially represent what is no longer physically present, transforming a sensory input into an internal cognitive token capable of interacting with the future unconditioned stimulus.

This representational requirement directly correlates with declarative awareness in humans. In seminal studies examining human trace eyeblink conditioning, successful acquisition was found to occur almost exclusively in participants who developed explicit, verbalizable awareness of the stimulus contingencies—specifically, the conscious realization that the offset of the tone predicted the subsequent delivery of the corneal airpuff. Participants who remained unaware of this temporal relationship, whether due to experimental masking or cognitive distraction, failed to reliably acquire conditioned responses. This stands in stark contrast to delay conditioning, which occurs independently of whether the subject can articulate the environmental contingencies.

Furthermore, trace conditioning demands dedicated top-down attentional allocation. Because the trace interval represents an environmental void, the central nervous system must actively shield the sustained neural representation from degradation by irrelevant ambient stimuli. The administration of concurrent secondary tasks that consume working memory capacity (such as continuous mental arithmetic or dual-task auditory discrimination) reliably interferes with and impairs trace conditioning acquisition. This top-down attentional modulation, orchestrated primarily by the prefrontal cortex, ensures that the organism remains focused on the internal temporal model of the environment until the anticipated US arrives.

3. Comprehensive Overview of Delay Conditioning

3.1 Temporal Architecture and Paradigmatic Structure

The temporal architecture of delay conditioning is defined by continuous, unbroken sensory overlap. In its standard formulation, the conditioned stimulus is initiated, persists through the entirety of the interstimulus interval, and overlaps directly with the unconditioned stimulus. In the most common experimental variant—coterminous delay conditioning—the CS terminates precisely at the moment the US ceases, meaning that both stimuli turn off simultaneously. In overlapping delay conditioning, the CS may persist throughout the US presentation and terminate shortly thereafter, or the US may be administered slightly prior to the termination of a prolonged CS. The common parameter across all delay paradigms is the complete absence of a temporal gap between CS offset and US onset.

Experimental protocols divide delay conditioning into two primary classifications: short-delay conditioning and long-delay conditioning. Short-delay conditioning typically employs an ISI ranging from 200 to 500 milliseconds (in motor reflex systems such as eyeblink conditioning). This range corresponds closely with the optimal physiological parameters for rapid associative plasticity in the cerebellum and autonomic brainstem pathways. In short-delay conditioning, CR acquisition is extraordinarily rapid, robust, and exhibits exceptional resilience to degradation, as the high temporal proximity and overlap facilitate direct associative bonding.

Long-delay conditioning, by contrast, extends the CS duration to several seconds or even minutes prior to the delivery of the US, while still maintaining unbroken CS continuity until US arrival. Interestingly, long-delay conditioning introduces a sophisticated behavioral adaptation known as “inhibition of delay.” In the early phases of long-delay training, organisms often exhibit conditioned responses immediately following CS onset. However, as training progresses, the onset of the CR gradually migrates forward in time, becoming inhibited during the early moments of the CS and deploying precisely before the anticipated arrival of the US. Although long-delay conditioning involves complex timing mechanisms that can occasionally recruit forebrain networks, its defining feature remains the uninterrupted physical presence of the CS up to the delivery of the reinforcer.

3.2 Neurobiological Substrates and Circuitry

The neural circuitry that supports delay conditioning represents one of the most thoroughly mapped associative learning networks in vertebrate neurobiology. Pioneered by Richard F. Thompson and colleagues through the rabbit nictitating membrane preparation, extensive research has established that the essential circuitry for the acquisition, consolidation, and expression of standard short-delay conditioning resides entirely within the cerebellum and its associated brainstem reflex pathways, operating independently of the cerebral cortex and hippocampus.

The canonical circuit for delay eyeblink conditioning operates through two convergent anatomical pathways that intersect within the cerebellum. The conditioned stimulus (e.g., an auditory tone) is transduced by peripheral receptors and relayed through sensory brainstem pathways, specifically the ventral cochlear nucleus, to the pontine nuclei. The pontine nuclei send excitatory mossy fibers into the cerebellum. These mossy fibers bifurcate: one branch provides direct collateral excitation to the deep cerebellar nuclei—specifically the anterior interpositus nucleus—while the other ascends into the cerebellar cortex to synapse upon granule cells. The granule cells, in turn, emit unmyelinated axons known as parallel fibers, which form hundreds of thousands of excitatory synapses upon the extensive dendritic trees of GABAergic Purkinje cells.

Concurrently, the unconditioned stimulus (e.g., a corneal airpuff or periorbital shock) activates sensory trigeminal pathways, which project directly to the dorsal accessory olive of the inferior olivary nucleus in the caudal medulla. The inferior olive emits climbing fibers that climb the dendritic arbors of Purkinje cells, delivering a massive, multi-spike excitatory complex. The climbing fiber pathway acts as a primary neurobiological “error signal” or reinforcement pathway. When the mossy fiber input (CS) and the climbing fiber input (US) converge upon the Purkinje cell dendrites within an optimal temporal window (100 to 500 milliseconds), associative synaptic plasticity is induced.

The primary form of plasticity in the cerebellar cortex is long-term depression (LTD) at the parallel fiber–Purkinje cell synapse. Purkinje cells constitutively exert powerful, continuous tonically inhibitory GABAergic control over the underlying deep cerebellar nuclei. When parallel fiber-Purkinje cell LTD is induced via metabotropic glutamate receptor (mGluR1) activation, intracellular calcium influx, and the clathrin-mediated endocytosis of post-synaptic AMPA receptors, the Purkinje cell fires fewer simple spikes in response to the CS. This reduction in Purkinje cell firing produces a release from inhibition—a disinhibition—of the anterior interpositus nucleus. The liberated interpositus neurons fire bursts of action potentials that travel via the superior cerebellar peduncle to the red nucleus in the midbrain, which then projects via the rubrobulbar tract to the facial motor nucleus (cranial nerve VII) and abducens nucleus (cranial nerve VI), driving the discrete, timely closure of the eyelid.

Parallel plasticity also occurs within the interpositus nucleus itself, where repetitive CS-US pairings induce long-term potentiation (LTP) at the mossy fiber-interpositus synapse. Because the basic sensory-motor loop (pontine-cerebellar-olivary-rubral) contains all necessary associative, timing, and motor machinery, standard short-delay conditioning remains completely unaffected by decerebration, total decortication, or complete bilateral ablation of the hippocampus.

3.3 Cognitive Characteristics and Processing Mode

Because its underlying physical circuitry is contained within subcortical and cerebellar loops, delay conditioning is characterized by an automatic, non-declarative, and implicit mode of cognitive processing. It functions as a classical model of procedural memory. Learning in standard delay paradigms unfolds inexorably across repeated trials without demanding conscious awareness of the predictive contingencies. The association is stamped into the nervous system as a direct, mechanistic adaptation to physical stimulus regularities, operating beneath the level of conscious executive processing.

This automaticity is evidenced by the resilience of delay conditioning to concurrent cognitive demands. Unlike higher-order cognitive functions that rely heavily on finite central executive bandwidth, delay conditioning acquisition rates remain stable when human participants are simultaneously engaged in demanding distracting tasks, such as continuous visual tracking, semantic classification, or working memory n-back challenges. The cognitive load that degrades performance on declarative memory tasks fails to penetrate or disrupt the localized cerebellar synaptic plasticity that drives delay learning.

Moreover, delay conditioning exhibits preservation across altered states of consciousness and profound neurological dissociation. Robust delay conditioning has been observed in animal models under deep pharmacological sedation and surgical decerebration. In human clinical populations, amnesic patients characterized by total destruction of the medial temporal lobes exhibit entirely normal acquisition rates, retention intervals, and extinction kinetics during short-delay eyeblink conditioning, despite retaining zero conscious episodic recollection of the experimental apparatus, the training trials, or the clinical personnel administering the test.

4. Key Differences Between Trace and Delay Conditioning

4.1 Temporal Disparity and Stimulus Overlap

The absolute operational distinction between trace and delay conditioning resides within the presence versus the complete absence of a stimulus-free temporal gap. In delay conditioning, associative linking is supported by continuous sensory reality. The physical presence of the conditioned stimulus bridges the interstimulus interval, ensuring that the primary sensory pathways conveying the CS remain active and electrophysiologically engaged at the precise moment the unconditioned stimulus reaches the central nervous system. The nervous system is presented with overlapping sensory activations, allowing temporal contiguity to be computed directly from peripheral inputs.

Trace conditioning, conversely, necessitates retrospective mnemonic association across an environmental void. Because the physical CS vanishes prior to US delivery, primary sensory reception pathways return toward baseline firing rates before the reinforcer arrives. The nervous system cannot execute an association between concurrent sensory streams. Instead, it must pair a present physical stimulus (the US) with an internal, post-sensory mnemonic representation of an event that occurred in the past (the trace of the CS). This simple operational gap of a few hundred milliseconds fundamentally transforms the computational problem from one of sensory integration into one of mnemonic maintenance and temporal retrieval.

This temporal disparity also dictates the paradigms’ differential vulnerability to variations in the interstimulus interval. Delay conditioning can tolerate moderate lengthening of the ISI (extending from 200 ms up to several seconds in long-delay variants) without demanding a categorical shift in the underlying anatomical structures, as the stimulus remains physically present. In trace conditioning, however, the rate of acquisition drops sharply as the trace interval is extended beyond approximately 500 to 1000 milliseconds in mammalian eyeblink paradigms. The biological capacity of cortical and hippocampal networks to sustain an active, high-fidelity neural trace without sensory input is inherently constrained by cellular energetics, intrinsic leak conductances, and network-level interference.

Head-to-Head Comparison: Trace vs. Delay Conditioning

Comparative Dimension	Trace Conditioning	Delay Conditioning
Temporal Configuration	CS offset precedes US onset; separated by an explicit, stimulus-free trace interval.	CS onset precedes US onset; CS persists through the ISI and overlaps/coterminates with the US.
Physical Gap	Present (temporal discontinuity requiring internal bridging).	Completely absent (zero temporal hiatus between stimuli).
Primary Neural Substrates	Hippocampus (CA1/CA3), Medial Prefrontal Cortex, Cerebellum, Brainstem nuclei.	Cerebellar cortex (Purkinje cells), Deep Cerebellar Nuclei (Interpositus), Pontine & Inferior Olive nuclei.
Forebrain Dependency	Obligate: Lesions to the hippocampus or mPFC abolish acquisition.	Independent: Preserved following decerebration, decortication, or hippocampectomy.
Cognitive Classification	Declarative / Representational memory system.	Non-declarative / Procedural memory system.
Contingency Awareness	Requires explicit, conscious awareness of the CS–US predictive relationship in humans.	Occurs automatically and implicitly without requiring conscious contingency awareness.
Working Memory Load	Highly vulnerable to secondary cognitive distractors and working memory interference.	Resilient and unaffected by concurrent cognitive demands or working memory distractors.
Synaptic Plasticity Sites	NMDA-dependent LTP in CA1/CA3; sustained prefrontal persistent firing; cerebellar LTD/LTP.	mGluR1/AMPA-receptor endocytosis (LTD) at parallel fiber–Purkinje synapses; interpositus LTP.
Optimal ISI Range (Eyeblink)	Tight physiological window; acquisition degrades sharply as trace interval exceeds 500–1000 ms.	Broad operational tolerance (200–500 ms optimal for short-delay; seconds for long-delay).
Clinical Vulnerability	Early biomarker for Alzheimer’s disease, MCI, medial temporal amnesia, and schizophrenia.	Selectively impaired in cerebellar ataxias, olivopontocerebellar atrophy; spared in pure amnesia.

4.2 Functional Neuroanatomy and Circuit Specialization

The anatomical divergence between trace and delay conditioning represents one of the clearest neuroanatomical double dissociations established in behavioral science. Delay conditioning relies upon a highly compact, specialized, subcortical-cerebellar reflex arc. Sensory mossy fibers carrying CS information from the basilar pontine nuclei converge with climbing fibers carrying US error signals from the inferior olivary nucleus directly onto Purkinje cells in the cerebellar cortex and projection neurons within the anterior interpositus nucleus. Synaptic plasticity at these localized sites—specifically parallel fiber-Purkinje cell LTD and mossy fiber-interpositus LTP—is fully sufficient to generate, time, and express the conditioned motor response.

Trace conditioning incorporates this exact cerebellar-brainstem motor execution circuit, but subordinates it to an upstream forebrain control network. Because the mossy fiber input from the pontine nuclei terminates when the physical CS turns off, the cerebellar cortex receives no sensory input during the trace interval. Consequently, the cerebellum cannot independently calculate the associative pairing or the temporal metric. The central nervous system circumvents this limitation by routing the sensory signal through the primary sensory cortices into the entorhinal-hippocampal loop and the medial prefrontal cortex.

Within this upstream network, persistent cellular firing across the CA1-CA3 networks and prefrontal pyramidal ensembles maintains the predictive signal across the temporal gap. At the end of the trace interval, this sustained forebrain representation drives descending efferent projections—relayed through the basal ganglia, thalamus, and direct prefrontal projections to the pontine nuclei—which reactivates the mossy fiber input to the cerebellum just prior to US arrival. Therefore, while delay conditioning requires only the cerebellum, trace conditioning requires the cerebellum plus the coordinated machinery of the telencephalon.

This functional specialization extends directly to divergent neurochemical and synaptic mechanisms. While delay conditioning depends critically on metabotropic glutamate receptor cascades (mGluR1), protein kinase C (PKC) activation, and the phosphorylation-dependent endocytosis of GluA2-containing AMPA receptors at the cerebellar Purkinje cell dendritic arbor, trace conditioning imposes an absolute requirement for forebrain N-methyl-D-aspartate (NMDA) receptor-mediated plasticity. Pharmacological blockade or genetic deletion of the GluN1 or GluN2B subunits of the NMDA receptor within the dorsal hippocampus completely prevents the acquisition of trace eyeblink conditioning, while leaving delay eyeblink conditioning entirely intact.

4.3 Awareness, Attention, and Cognitive Demands

The psychological constructs of awareness and selective attention draw a sharp line between trace and delay conditioning. In a landmark series of human psychophysiological experiments conducted by Robert E. Clark and Larry R. Squire in 1998, the role of conscious contingency awareness was directly evaluated using post-conditioning structured debriefing questionnaires. The findings revealed that in trace eyeblink paradigms (using a 280-millisecond tone followed by an empty 500-millisecond trace interval), conditioned responses emerged exclusively in participants who were able to explicitly articulate that the tone preceded and predicted the airpuff. Individuals who remained unaware of the relationship performed at chance levels, exhibiting no conditioned blinking.

In contrast, during delay conditioning (where the tone persisted for 780 milliseconds and coterminated with the airpuff), both aware and completely unaware participants acquired robust conditioned responses at identical rates and to equivalent asymptotic levels. This demonstrated that conscious awareness is not an epiphenomenon of conditioning in general, but is an obligate operational requirement for the bridge mechanism in trace paradigms. Conscious awareness reflects the recruitment of the declarative memory system, mediated by the medial temporal lobe, which acts to deliberately bridge the temporal hiatus.

Because trace conditioning is intimately tied to declarative processing, it is uniquely susceptible to disruption by concurrent cognitive demands. When subjects are subjected to dual-task paradigms—such as performing a continuous visual serial-reaction-time task, tracking complex visual stimuli, or executing working memory updates—trace conditioning is severely degraded or entirely blocked. The secondary task monopolizes prefrontal executive resources and working memory storage, depriving the hippocampal-prefrontal network of the bandwidth needed to sustain the active CS trace. Delay conditioning, running on autonomous cerebellar circuitry, proceeds unaffected under identical cognitive dual-task burdens.

4.4 Pathological Vulnerability and Clinical Sensitivity

Due to their divergent neuroanatomical foundations, trace and delay conditioning exhibit distinct patterns of susceptibility across clinical neuropathologies, providing profound utility in differential diagnosis. Patients suffering from medial temporal lobe amnesia—such as the famous case of patient H.M., or individuals who have sustained bilateral ischemic damage to the CA1 field of the hippocampus—present with normal delay conditioning. They acquire, consolidate, and extinguish delay eyeblink responses with healthy trajectories. Yet, these identical individuals are completely incapable of acquiring trace conditioning when a stimulus-free interval of 500 milliseconds or longer is introduced.

This double dissociation is highlighted in neurodegenerative conditions, notably Alzheimer’s disease (AD) and its precursor, Amnestic Mild Cognitive Impairment (aMCI). The earliest neuropathological hallmarks of Alzheimer’s disease—specifically neurofibrillary tangle deposition and synaptic loss—occur within the transentorhinal cortex, entorhinal cortex, and the CA1 subfield of the hippocampus. Consequently, trace conditioning deficits emerge during the preclinical stages of AD, long before overt deficits can be tracked on generalized neuropsychological screening instruments. In contrast, cerebellar integrity is preserved until the terminal stages of AD, meaning that delay conditioning remains intact throughout the majority of the disease course.

Conversely, pathologies that target the hindbrain selectively disrupt delay conditioning. Patients with pure cerebellar cortical degeneration, spinocerebellar ataxias, or localized ischemic strokes within the distribution of the superior cerebellar artery exhibiting lesions to the anterior interpositus nucleus show devastating deficits in delay conditioning. They are completely unable to acquire conditioned motor responses, or produce conditioned responses that are severely mistimed (e.g., executing the blink after the airpuff has already struck the eye). If their forebrain structures remain intact, their declarative awareness of the contingency persists unhindered: they can verbally report that the tone signals the airpuff, yet their motor system cannot instantiate the reflexive associative motor memory.

In psychiatric conditions such as schizophrenia, trace conditioning exhibits profound selective impairment that correlates directly with prefrontal cortical dysfunction and aberrant hippocampal-prefrontal theta synchrony. Schizophrenic patients routinely demonstrate marked deficits in trace conditioning acquisition, attributed to hypofunction of prefrontal NMDA receptors and disrupted dopaminergic neuromodulation, despite presenting with preserved cerebellar delay conditioning. In neurodevelopmental disorders like Autism Spectrum Disorder (ASD), subtle abnormalities in cerebellar Purkinje cell density often induce idiosyncratic shifts in delay conditioning timing kinetics (e.g., abnormally accelerated acquisition with premature CR peak latencies), while trace conditioning may remain comparatively preserved or display distinct impairments related to frontostriatal connectivity.

5. Similarities and Theoretical Overlap

5.1 Shared Learning Dynamics and Behavioral Curves

Despite their deep anatomical and cognitive differences, trace and delay conditioning share striking commonalities in their overarching learning dynamics and macroscopic behavioral profiles. When plotted across successful training sessions, both paradigms yield the classical sigmoidal (S-shaped) learning curve characteristic of basic associative learning. Both paradigms typically exhibit an initial latency phase (where associative strength is quietly accruing beneath the behavioral expression threshold), followed by a steep acquisition phase of accelerating CR frequency, ultimately leveling off into an asymptotic plateau representing maximum associative strength.

Furthermore, both paradigms are fully subject to the rules of experimental extinction. If the conditioned stimulus—whether presented continuously as in delay or briefly followed by a gap as in trace—is repeatedly administered in the absolute absence of the unconditioned stimulus, the conditioned response frequency systematically declines until returning to baseline pre-conditioning levels. In both paradigms, extinction does not represent the passive unlearning or erasure of the original engram, but constitutes an active, inhibitory learning process. This is confirmed by the universal manifestation of post-extinction recovery phenomena across both trace and delay paradigms:

• Spontaneous Recovery: The spontaneous reappearance of the extinguished conditioned response following the mere passage of time without any further CS-US pairings.
• Renewal: The immediate recovery of the extinguished CR when the CS is tested in a physical or sensory context different from the context in which extinction training occurred (e.g., ABA renewal).
• Reinstatement: The rapid restoration of the extinguished conditioned response produced by the isolated, non-contingent presentation of the unconditioned stimulus alone.

5.2 General Classical Conditioning Principles

Both trace and delay conditioning are governed by the established mathematical and computational frameworks of associative learning, most prominently the Rescorla-Wagner model. Formulated in 1972, this predictive model dictates that the change in associative strength ($\Delta V$) that occurs on any given conditioning trial is a direct mathematical function of prediction error—the discrepancy between the actual biological value of the unconditioned stimulus ($lambda$) and the aggregate associative strength of all predictive cues present on that trial ($\sum V$):

$$\Delta V = \alpha \beta (\lambda – \sum V)$$

Where $\alpha$ represents the salience of the conditioned stimulus, and $\beta$ represents the learning-rate parameter of the unconditioned stimulus. Both trace and delay conditioning rigorously conform to these mathematical parameters. If a conditioned stimulus provides no new informational utility or predictive value, associative strength fails to accrue, regardless of physical timing parameters.

Consequently, both paradigms display the classical cue-competition phenomena derived from prediction error computations, including:

• Blocking: Prior successful conditioning to stimulus A completely blocks the acquisition of associative strength to a novel stimulus B when they are subsequently presented together as a compound cue (AB) paired with the US.
• Overshadowing: When two conditioned stimuli of differential physical salience (e.g., a high-decibel auditory tone paired with a low-intensity visual light) are presented as a compound cue, the high-salience stimulus captures the vast majority of associative strength, impairing learning to the less salient stimulus.
• Statistical Contingency: Both paradigms require true statistical contingency ($\Delta P = P(\text{US}|\text{CS}) – P(\text{US}|\text{no CS}) > 0$). Mere temporal contiguity alone is completely insufficient to support conditioning if non-contingent background presentations of the US degrade the predictive statistical relationship between the CS and the reinforcer.

5.3 Overlap in Effector Systems and Motor Execution

The ultimate execution of the behavioral conditioned response converges upon identical motor pathways, irrespective of whether the upstream computation was driven by a delay or a trace architecture. In the somatic motor domain (exemplified by the classic eyeblink or nictitating membrane preparation), both trace and delay conditioning rely on the anterior interpositus nucleus of the cerebellum as the final common integration center for executing motor timing commands. From the interpositus, the efferent motor commands traverse the superior cerebellar peduncle to the magnocellular red nucleus, descend via the rubrobulbar tract, and terminate directly upon the motor nuclei of cranial nerves VI (abducens) and VII (facial), which innervate the orbicularis oculi and retractor bulbi muscles to physically close the eyelid.

Similarly, when examining autonomic and emotional conditioning paradigms (such as conditioned fear assays), both trace and delay paradigms converge upon the central nucleus of the amygdala (CeA) and the lateral amygdala (LA). The execution of conditioned fear responses—including behavioral freezing, fear-potentiated startle, changes in respiration, and elevations in galvanic skin response (GSR) or skin conductance—is mediated through downstream amygdalar projections to the ventrolateral periaqueductal gray (vlPAG) for somatic freezing and the lateral hypothalamus for autonomic sympathetic mobilization.

Most remarkably, both paradigms exhibit an identical standard of temporal precision in the motor execution of the conditioned response. The conditioned response is never an uncoordinated, immediate spasm triggered indiscriminately by the CS. Instead, the central nervous system calculates the temporal delay such that the peak amplitude of the conditioned response coincides with the arrival of the unconditioned stimulus. Whether the CS persists throughout the interval (delay) or terminates early (trace), the CR emerges gradually, building smoothly to achieve maximum eyelid closure or maximum autonomic readiness at the moment the airpuff or footshock arrives, protecting the peripheral tissues from physical trauma.

6. Practical Experimental Applications and Paradigm Selection

6.1 Experimental Eyeblink Conditioning Paradigms

The standard somatic assay for evaluating classical conditioning across mammalian species is the eyeblink conditioning (EBC) preparation, utilized extensively in rodents, rabbits, and human participants. In the typical experimental setup, the subject is positioned in an environment where a conditioned stimulus—usually a 1 kHz to 2 kHz auditory pure tone delivered via headphones or speakers at approximately 75 to 85 decibels—is paired with an unconditioned stimulus that reliably produces a corneal blink. The US typically consists of a brief, localized puff of compressed medical-grade air (e.g., 3 to 5 psi, 50 to 100 ms duration) directed via an adjustable nozzle toward the cornea, or a mild transcutaneous electrical micro-stimulation applied to the supraorbital branch of the trigeminal nerve.

Conditioned responses are measured via high-resolution physiological instrumentation. In animal models, fine electromyographic (EMG) wire electrodes are surgically implanted into the orbicularis oculi muscle of the upper eyelid to capture real-time muscle action potentials, or an infrared reflectance sensor is mounted on the skull assembly to track eyelid displacement without direct mechanical contact. In human investigations, surface EMG electrodes are placed infraorbitally over the orbicularis oculi, or an infrared photo-transistor mounted to an experimental headset measures reflected infrared light from the ocular surface, yielding microsecond-level temporal resolution of eyelid closure.

Researchers strategically deploy delay eyeblink conditioning when the explicit scientific goal is to isolate and interrogate the functional integrity of the cerebellar-brainstem network. Because delay EBC is independent of the cerebral cortex and hippocampus, it serves as an exquisite clinical and basic science assay to evaluate cerebellar plasticity, the effects of localized cerebellar cerebellar strokes, spinocerebellar ataxias, or the pharmacological modulation of Purkinje cell AMPA-receptor trafficking. Conversely, trace eyeblink conditioning is specifically deployed to evaluate the integrity of forebrain networks, long-range hippocampal-prefrontal oscillatory synchronization, and declarative memory acquisition. By comparing the performance of a single cohort across both trace and delay eyeblink protocols, investigators can determine whether an observed behavioral deficit stems from basic sensorimotor execution failures within the cerebellum or higher-order representational deficits within the medial temporal lobe.

6.2 Conditioned Fear and Autonomic Response Assays

In addition to discrete somatic motor preparations, classical conditioning paradigms are standardly implemented in fear and defensive behavior research. In rodent auditory fear conditioning, an auditory conditioned stimulus is paired with an aversive unconditioned stimulus, specifically a mild, unescapable transcutaneous footshock (e.g., 0.4 to 0.8 mA, 0.5 to 1.0-second duration) delivered through an electrified stainless steel grid floor. Conditioning is indexed via automated video-tracking systems that quantify behavioral freezing—defined as the complete absence of all movement except for respiratory motions—or via accelerometer-equipped cages measuring fear-potentiated startle responses.

In delay fear conditioning, the tone sounds continuously for 20 to 30 seconds and coterminates with the presentation of the footshock. This form of conditioning is critically dependent upon the basolateral amygdala complex (BLA), requiring zero functional contribution from the dorsal hippocampus. The sensory inputs for both the continuous auditory CS (relayed via the auditory thalamus and auditory cortex) and the somatosensory shock US (relayed via the spinothalamic tract) converge directly onto pyramidal projection neurons within the lateral nucleus of the amygdala, driving long-term potentiation of thalamo-amygdala synapses.

In trace fear conditioning, an empty trace interval (ranging from 15 to 30 seconds) separates the offset of the auditory CS from the onset of the footshock. The inclusion of this trace interval recruits the dorsal hippocampus, the ventral hippocampus, the prelimbic cortex, and the anterior cingulate cortex. Lesions or pharmacological inactivation of the dorsal hippocampus immediately before or shortly after training eliminate trace fear retention while leaving delay fear conditioning preserved. In human cognitive neuroscience, fear paradigms standardly measure autonomic sympathetic responses via galvanic skin response (GSR) or skin conductance responses (SCR) using silver/silver-chloride electrodes placed upon the hypothenar or thenar eminences of the hand. Human GSR trace fear paradigms demonstrate identical prefrontal-hippocampal dependencies and parallel requirements for conscious contingency awareness, making them invaluable for cross-species translational psychiatric research.

6.3 Strategic Guidelines for Research Paradigm Selection

Selecting the optimal temporal conditioning paradigm requires a clear alignment between the experimental hypothesis, the target neural circuits, and the underlying cognitive constructs under investigation. The strategic decision process should adhere to the following experimental guidelines:

• Choose Delay Conditioning when:
  • The primary objective is to evaluate basic associative motor learning, synaptic plasticity in the cerebellum (e.g., Purkinje cell LTD), or brainstem reflex physiology.
  • The experimental design requires testing in subjects with compromised cognitive, forebrain, or attentional capacities (e.g., deeply sedated models, decerebrate preparations, or patients with severe medial temporal lobe amnesia).
  • The research question demands an implicit, automatic learning metric that is resilient to cognitive dual-task interference, mind-wandering, or differences in participant intelligence and working memory capacity.
• Choose Trace Conditioning when:
  • The experimental objective is to assess the functional connectivity and integrity of the medial temporal lobe, the hippocampus, and the medial prefrontal cortex.
  • The study aims to investigate the mechanisms of working memory maintenance, temporal integration, and the neural substrates that bridge physical stimulus gaps.
  • The investigator seeks a sensitive behavioral assay for early-stage neurodegenerative disorders (e.g., Alzheimer’s disease, aMCI) or psychiatric disorders characterized by prefrontal and oscillatory dysregulation (e.g., schizophrenia).
  • The research explores the boundary conditions between declarative, conscious awareness and non-declarative, procedural memory systems.

When implementing these paradigms, investigators must rigorously control for critical confounding variables. Chief among these is the physical duration of the conditioned stimulus. If a researcher compares a delay protocol with an ISI of 500 ms (CS duration of 500 ms) against a trace protocol with an ISI of 500 ms (CS duration of 100 ms, trace interval of 400 ms), the total sensory energy delivered by the CS differs five-fold. To establish true empirical parity, controls must verify that observed performance deficits in the trace group are not simply artifacts of reduced sensory salience or insufficient sensory integration time. Modality-specific sensory persistence, ambient acoustic noise, experimental context shifts, and differential unconditioned arousal effects must also be systematically calibrated to ensure valid, reproducible outcomes.

7. Frequently Asked Questions

7.1 Can Delay Conditioning Occur Without Conscious Awareness?

Yes, robust empirical evidence demonstrates that delay conditioning occurs entirely in the absence of conscious awareness of the predictive contingencies. In human studies utilizing subliminal visual stimuli, masked conditioned stimuli, or auditory signals presented under conditions of high cognitive distraction, subjects reliably acquire normal conditioned eyeblink and autonomic responses. The subject’s eyelid blinks at the precise arrival of the airpuff, yet when questioned, the individual is incapable of reporting that the cue predicted the reinforcer, or that conditioning even occurred. This automaticity is supported by non-human animal preparations involving total surgical decerebration or profound general anesthesia, wherein localized synaptic plasticity within the cerebellar cortex and brainstem continues to instantiate delay conditioning. As a non-declarative, procedural learning mechanism mediated by subcortical circuits, delay conditioning operates below the threshold of subjective conscious perception.

7.2 Why Does Trace Conditioning Specifically Require the Hippocampus?

Trace conditioning specifically requires the hippocampus because the physical conditioned stimulus terminates before the unconditioned stimulus arrives, leaving an environmental void that primary sensory and cerebellar circuits cannot autonomously bridge. The basilar pontine nuclei cease firing when the CS turns off; therefore, the cerebellum receives no direct sensory input during the trace interval. The hippocampus—specifically the CA1 and CA3 pyramidal fields—possesses unique cellular and network adaptations, including persistent “time-cell” firing, low rates of synaptic decay, and intrinsic non-specific cation currents supported by cholinergic inputs, that allow it to sustain an active internal representation of the absent CS across time. Coordinated via reciprocal theta-band (4–8 Hz) oscillations with the medial prefrontal cortex, the hippocampus maintains this neural representation and provides the descending drive back to the cerebellar and brainstem motor systems at the conclusion of the trace interval, allowing the motor apparatus to execute an accurately timed association.

7.3 How Does Trace Interval Length Influence Associative Strength?

The length of the trace interval exhibits a strict, inverse relationship with associative acquisition speed and terminal response strength. In classical eyeblink paradigms, as the trace interval is extended from 250 milliseconds to 500 milliseconds, the number of trials required to acquire conditioned responses increases significantly. When the trace interval exceeds approximately 1000 to 2000 milliseconds in standard mammalian eyeblink conditioning, successful behavioral acquisition typically fails to occur without specialized modifications. This degradation occurs because sustaining persistent neuronal activity across an empty temporal interval requires significant energetic resources and is inherently vulnerable to endogenous neurochemical decay and network interference. In laboratory settings, researchers can extend this operational window by introducing environmental bridging strategies, increasing CS salience, or providing explicit top-down prefrontal attentional priming.

7.4 Can a Trace Paradigm Functionally Shift to a Delay Paradigm?

Under specific physiological and experimental conditions, an intended trace paradigm can functionally collapse into a delay-like paradigm if a secondary physical substrate bridges the stimulus-free gap. The most frequent source of this artifact is sensory persistence within the peripheral nervous system. If an auditory CS is delivered at high intensity, mechanical reverberations within the cochlea or persistent sensory firing in primary acoustic pathways can endure for hundreds of milliseconds after physical tone offset, effectively delivering continuous sensory input to the brainstem. Similarly, high-intensity visual cues can produce retinal after-images that persist across the nominal trace interval. Furthermore, the experimental apparatus and environmental background—the physical context—can inadvertently become conditioned, serving as a continuous physical CS that overlaps directly with the US. To prevent these experimental artifacts, researchers must use calibrated, moderate stimulus intensities and implement explicit context-extinction pre-exposure phases to preserve the operational integrity of the trace interval.

8. Conclusion

8.1 Synthesis of Temporal Architecture and Neural Segregation

The neurobiological divergence between trace and delay conditioning demonstrates how discrete physical parameters alter the cognitive and physiological demands of associative learning. The fundamental variable that separates these two archetypes is not simply the chronological duration between stimulus onsets, but the operational presence or absence of sensory continuity. In delay conditioning, the unconditioned stimulus is delivered into a nervous system whose sensory reception channels are actively engaged with the conditioned stimulus. This unbroken temporal overlap allows the learning event to be resolved entirely within autonomous subcortical and cerebellar circuits, operating as an implicit, procedural reflex adaptation independent of conscious awareness or higher-order executive control.

In trace conditioning, the absolute termination of the CS prior to US delivery breaks this sensory continuity, confronting the central nervous system with an environmental void. To resolve this temporal hiatus, the computational burden shifts upward to the forebrain. The organism must generate an internal, mnemonic token of the past event, buffer that representation within fronto-hippocampal working memory networks, recruit explicit attentional resources, and actively project a timing signal downward to the cerebellar motor execution machinery. Consequently, trace conditioning functions as a representational, declarative learning process. The contrast between these paradigms provides empirical validation for dual-process theories of memory, demonstrating that distinct subcortical and telencephalic memory systems operate concurrently, recruited dynamically based on the temporal architecture of environmental inputs.

8.2 Future Trajectories in Temporal Conditioning Research

As behavioral neuroscience advances into an era defined by high-density electrophysiology, deep-brain imaging, and cell-type-specific genetic manipulations, the comparative study of trace and delay conditioning continues to yield transformative insights. The deployment of optogenetics and chemogenetics (DREADDs) is allowing investigators to selectively silence or activate individual “time-cell” ensembles within the CA1 field of the hippocampus or specific projection neurons within the prelimbic-pontine pathway with millisecond precision, dissecting the exact synaptic mechanisms that sustain mnemonic traces across temporal intervals.

Simultaneously, computational neuroscience is utilizing these empirical data to refine continuous-time temporal difference (TD) reinforcement learning models and internal clock algorithms. By mapping the neuroanatomical divergence between cerebellar parallel fiber LTD and hippocampal-prefrontal theta synchrony, computational theorists are constructing biologically realistic neural network architectures that unify procedural motor timing with higher-order cognitive episodic anticipation. In the clinical sphere, automated, high-throughput digital eyeblink conditioning assays are emerging as promising, cost-effective biomarkers for the early screening of neurodegenerative diseases. By exploiting the vulnerability of trace conditioning to early entorhinal and hippocampal degradation, clinicians can track subclinical cognitive decline long before the emergence of generalized dementia. Thus, Pavlov’s classic temporal paradigms remain foundational instruments in deciphering the neural code that binds memory across time.

References

• Clark, R. E., & Squire, L. R. (1998). Classical conditioning and awareness. Science, 280(5360), 77–81. https://doi.org/10.1126/science.280.5360.77
• Moyer, J. R., Jr., Deyo, R. A., & Disterhoft, J. F. (1990). Hippocampectomy impairs trace eyeblink conditioning in rabbits without affecting delay conditioning. Behavioral Neuroscience, 104(2), 243–252. https://doi.org/10.1037/0735-7044.104.2.243
• Pavlov, I. P. (1927). Conditioned Reflexes: An Investigation of the Physiological Activity of the Cerebral Cortex (G. V. Anrep, Trans.). Oxford University Press.
• Rescorla, R. A., & Wagner, A. R. (1972). A theory of Pavlovian conditioning: Variations in the effectiveness of reinforcement and nonreinforcement. In A. H. Black & W. F. Prokasy (Eds.), Classical Conditioning II: Current Research and Theory (pp. 64–99). Appleton-Century-Crofts.
• Solomon, P. R., Vander Schaaf, E. R., Thompson, R. F., & Weisz, D. J. (1986). Hippocampus and trace conditioning of the rabbit’s classically conditioned nictitating membrane response. Behavioral Neuroscience, 100(5), 729–744. https://doi.org/10.1037/0735-7044.100.5.729
• Thompson, R. F. (1986). The neurobiology of learning and memory. Science, 233(4767), 941–947. https://doi.org/10.1126/science.3738519