The Evaluative Conditioning Experiment – Irene Martin and A.B. Levey

The landscape of mid-twentieth-century behavioral psychology was anchored to a fundamental premise: that associative learning is an adaptive mechanism designed to help organisms anticipate environmental events and execute preparatory motor or autonomic reflexes. From Ivan Pavlov’s salivating dogs to the precise, microsecond calibration of human eyeblink conditioning pioneered in the mid-century laboratories of Clark Hull and Kenneth Spence, conditioned responses were conceptualized almost exclusively as functional, predictive motor or physiological adjustments. Within this rigid framework, learning was defined by contingency, temporal precedence, and survival utility. Yet this mechanistic orthodoxy left a glaring theoretical void. It failed to account for how human beings acquire their nuanced, idiosyncratic, and ubiquitous networks of subjective preference—the quiet, enduring judgments of “like” and “dislike,” attraction and revulsion, aesthetic pleasure and distaste that govern human daily decision-making.

In 1975, at the Institute of Psychiatry in London, two experimental psychologists—Irene Martin and A.B. Levey—published a deceptively modest empirical investigation titled “Evaluative Conditioning” in the journal Behaviour Research and Therapy. Working in the clinical and psychophysiological traditions established by Hans Eysenck at the Maudsley Hospital, Martin and Levey bypassed traditional autonomic effectors such as galvanic skin responses, pupillary dilation, or striate muscle contractions. Instead, they asked an entirely novel question: Can the fundamental hedonic valence of an ostensibly neutral stimulus be systematically altered merely by pairing it contiguously with an emotionally valenced stimulus, completely divorced from motor preparation or explicit survival drives? Utilizing a visual paradigm of fine art postcards categorized by participants into liked, disliked, and neutral sets, Martin and Levey demonstrated that hedonic valence itself could be conditioned.

The publication of the Martin and Levey paradigm marked the birth of Evaluative Conditioning (EC) as an independent empirical construct and ignited an intellectual revolution across learning theory, cognitive psychology, and social science. Over the ensuing half-century, their initial findings sparked fierce debates concerning the nature of human associative memory, the necessity of conscious contingency awareness, the structural divergence between preparatory reflex conditioning and evaluative learning, and the mechanisms underlying attitude formation. Today, evaluative conditioning represents a foundational pillar in our understanding of consumer behavior, implicit bias, clinical phobias, addiction, and affective neuroscience. This comprehensive treatise reconstructs the empirical, methodological, and conceptual architecture of Martin and Levey’s pioneering experiments, tracing the evolution of evaluative conditioning from a localized laboratory curiosity into an enduring paradigm of modern psychological science.

1. Historical Foundations and Conceptual Origins of Evaluative Conditioning

1.1 The Pre-1975 Behaviorist Paradigm and Affective Gaps

In the decades preceding Martin and Levey’s 1975 breakthrough, experimental psychology was dominated by theoretical frameworks that prioritized overt, quantifiable behavior over internal qualitative experience. The intellectual hegemony of Clark Hull‘s drive-reduction formulation and B.F. Skinner‘s radical operant behaviorism cast associative learning as a strictly functional enterprise. Within the Hullian paradigm, conditioned stimuli served as predictive environmental cues that initiated preparatory behaviors designed to satisfy primary biological drives such as hunger, thirst, or avoidance of tissue damage. The animal or human subject was viewed as an energetic calculating system, constantly optimizing survival odds by detecting statistical regularities in the physical environment.

Traditional Pavlovian paradigms mirrored this mechanistic teleology. Whether examining the classic canine digestive preparations of Ivan Pavlov or the meticulously recorded human eyeblink conditioning protocols championed by Kenneth Spence, researchers measured somatic and autonomic effectors that prepared the organism for an imminent unconditioned stimulus (US). The conditioned response (CR) was interpreted as an adaptive, preparatory motor adjustment. In eyeblink conditioning, the eyelid closes milliseconds before the delivery of an aversive corneal airpuff to minimize ocular trauma. In autonomic electrodermal conditioning, the skin conductance response reflects sympathetic nervous system arousal that readies the organism for fight or flight. Within these classical architectures, the conditioned response was understood to be functionally homologous or compensatory to the unconditioned response, fundamentally rooted in biological survival.

This orthodox architecture suffered from a profound affective blind spot. While it effectively explained how an organism learns that a tone signals an electric shock, it offered virtually no conceptual apparatus to explain how that tone might acquire a persistent, qualitative aura of “badness” or “unpleasantness” that outlasts its predictive utility. Early psychophysiological researchers who ventured into hedonic territory frequently conflated physiological arousal with qualitative valence. They assumed that monitoring peripheral autonomic parameters, such as heart rate deceleration or digital vasoconstriction, captured the totality of affective acquisition. However, these physiological indicators failed to map reliably onto subjective human verbal evaluations. A subject might demonstrate equivalent autonomic deceleration to an alarming sound, an aesthetically striking painting, or an unexpected cognitive query, yet the subjective hedonic quality—the felt valence—of those experiences remained wildly divergent. The traditional behaviorist paradigm possessed no systematic empirical vehicle to bridge the chasm between raw autonomic reflexivity and the fine-grained landscape of human subjective preference.

1.2 Emergence of the Evaluative Construct in Conditioning Research

The gradual emergence of an autonomous evaluative construct within associative learning theory gained significant momentum from psycholinguistic and psychometric advances in the mid-twentieth century, most notably through the work of Charles E. Osgood and his collaborators. Osgood’s development of the semantic differential technique revealed that across disparate human cultures, languages, and contexts, subjective meaning consistently organizes around three primary, orthogonal dimensions: Evaluation (good–bad), Potency (strong–weak), and Activity (active–passive). Crucially, the Evaluative factor accounted for the vast majority of the variance in human semantic attribution. Osgood, Suci, and Tannenbaum (1957) demonstrated that words and symbols carry an intrinsic hedonic charge that operates semi-independently of their physical referents, inspiring early semantic conditioning experiments by researchers like Arthur and Carolyn Staats.

In the late 1950s and 1960s, Arthur Staats and his colleagues attempted to condition word meaning by pairing visually presented neutral nonsense syllables (such as YOF or LAQ) with verbally presented words possessing positive or negative evaluative connotations (such as beauty, gift, or bitter, ugly). While these early semantic studies showed that evaluative ratings could shift following contiguous exposure, the mainstream learning community viewed them with intense skepticism. Methodological critics contended that Staats’ results were contaminated by semantic satiation, deliberate verbal mediation, and overt experimental demand characteristics. Because the stimuli were linguistic tokens processed through complex pre-existing lexical networks, critics argued that these findings did not reflect fundamental associative conditioning, but rather the conscious, deductive compliance of human subjects deciphering the experimenter’s linguistic puzzle.

It was precisely at this theoretical impasse that Irene Martin and A.B. Levey formulated their radical hypothesis. They recognized that subjective preference was not merely an ancillary epiphenomenon of cognitive processing or verbal compliance, nor was it reducible to an anticipatory physiological reflex. Instead, they hypothesized that the subjective valence of a stimulus—its core position on a continuous spectrum of liking versus disliking—could be transferred directly through stimulus-stimulus (S-S) contiguity. Martin and Levey sought to strip away the confounds of linguistic complexity and semantic mediation by using visual, non-verbal stimuli. They hypothesized that pairing a completely neutral visual image with a deeply liked or disliked visual image would result in a direct, automatic transfer of hedonic quality. In doing so, they formally conceptualized evaluative conditioning not as a linguistic artifact, but as a distinct, primary sub-variety of classical learning that operated via its own psychological rules.

1.3 Defining Evaluative Conditioning: The Martin-Levey Perspective

From the outset, Martin and Levey defined evaluative conditioning with conceptual precision: Evaluative conditioning is the change in the valence of a stimulus that results from pairing that stimulus with other stimuli that are systematically liked or disliked. In this formulation, the initially neutral item is designated as the conditioned stimulus (CS), the inherently valenced item serves as the unconditioned stimulus (US), and the acquired subjective shift in liking or disliking constitutes the conditioned response (CR). While this nomenclature borrowed directly from Pavlovian terminology, Martin and Levey were careful to establish that the nature of the evaluative CR was categorically distinct from the somatic and autonomic CRs of orthodox reflexology.

The core theoretical differentiation lay in the distinction between biological preparatory reflexes and qualitative affective states. In standard classical conditioning, the CR is prospective, directional, and preparatory: an animal salivates to anticipate food ingestion, or a human blinks to cushion an anticipated mechanical impact. The CR is functionally organized to optimize interactions with an imminent physical event. In stark contrast, Martin and Levey argued that the evaluative CR possesses no anticipatory or preparatory motor function. An acquired dislike for a visual stimulus does not prime the subject to avoid a physical impact, nor does an acquired liking prepare the gastrointestinal tract for metabolic digestion. Instead, evaluative conditioning represents a qualitative transformation of internal representation: the CS ceases to be perceived as neutral and comes to be perceived as inherently pleasant or unpleasant.

The epistemological significance of Martin and Levey’s work was profound. By treating subjective verbal evaluations as valid, reliable empirical endpoints rather than unscientific introspection, they legitimized the study of affect within experimental learning theory. They operationalized subjective feeling states using rigorous psychophysical scaling, demonstrating that human preference obeys systematic empirical laws. However, this pioneering approach immediately encountered theoretical skepticism from orthodox learning theorists. Skeptics argued that human affective memory structures are inherently declarative and cognitive, insisting that any associative change in subjective liking must be mediated by conscious propositional beliefs. Martin and Levey anticipated these critiques, contending that evaluative associations could be forged directly at an implicit, non-declarative level of processing—a revolutionary stance that challenged decades of behaviorist doctrine and laid the groundwork for contemporary dual-process models of human cognition.

2. Biographical and Theoretical Background of Irene Martin and A.B. Levey

2.1 Academic Context and Laboratory Affiliations

The development of evaluative conditioning occurred within one of the twentieth century’s most prestigious and intellectually rigorous psychiatric research institutions: the Institute of Psychiatry at the University of London, situated adjacent to the historic Maudsley Hospital. During the 1960s and 1970s, the Institute served as a primary global hub for experimental psychopathology and physiological psychology, largely under the formidable directorship of Hans J. Eysenck. Eysenck’s department was built upon an uncompromising commitment to empirical measurement, quantitative psychometrics, and the biological reduction of neurotic and personality variations to basic conditioning parameters.

Within this rigorous scientific milieu, Irene Martin and A.B. Levey established a highly productive research partnership. Martin was an accomplished psychophysiologist whose early career had focused on the fine-grained electromyographic and autonomic tracking of human emotional and motor responding. Levey brought an exceptional command of quantitative psychophysics, classical learning paradigms, and behavioral instrumentation. Together, they operated within an interdisciplinary nexus where clinical inquiries into neurotic anxiety, phobias, and personality traits merged directly with pure laboratory experimentation. Their work was profoundly shaped by the broader Maudsley ethos: if an internal psychological phenomenon exists, it must be capable of objective, laboratory-controlled operationalization and psychophysiological dissection.

A central pillar of the Eysenckian research tradition was the hypothesis that individual differences in personality—specifically introversion versus extraversion and emotional stability versus neuroticism—were rooted in differential conditionability. Eysenck postulated that introverts, possessing higher baseline levels of cortical arousal, would acquire conditioned reflexes more rapidly and resist extinction longer than extraverts. To test these theories, the Maudsley laboratories constructed some of the most technologically sophisticated human classical conditioning apparatuses in the world, specializing in human eyelid conditioning. It was within this environment of absolute precision, microsecond timing, and relentless measurement of conditioning parameters that Martin and Levey cultivated their experimental methodology, an ethos they would soon apply directly to the ephemeral domain of human aesthetic preference.

2.2 Theoretical Trajectory: From Eyelid Conditioning to Subjective Valence

Martin and Levey’s pivot toward evaluative conditioning was an organic evolution originating within their empirical work on the human conditioned eyeblink response. Throughout the late 1960s and early 1970s, the duo published extensive monographs and experimental investigations detailing the parameters of human eyelid conditioning. Their laboratory utilized high-precision pneumatic delivery systems that administered micro-blasts of compressed nitrogen to the cornea, tracking reflexive and conditioned blinks via delicate photo-electric cells and electromyographic (EMG) sensors mounted on customized dental-impression head rigs. This work demanded extraordinary control over environmental noise, postural sway, and somatic artifacts.

However, as they immersed themselves in the micro-analysis of conditioned eyelid responses, Martin and Levey observed a persistent, anomalies-laden pattern that traditional conditioning models struggled to explain. While the mechanical acquisition of the somatic blink response adhered strictly to narrow inter-stimulus intervals (typically peaking at a precise 400 to 800 millisecond CS-US onset asynchrony), participants consistently reported qualitative psychological reactions to the experimental stimuli that operated completely outside these strict somatic temporal windows. Participants frequently developed intense, generalized aversions to the neutral auditory tones or visual warning cues used as conditioned stimuli. These subjective emotional orientations persisted even when the motor blink response failed to acquire, or long after the motor response had extinguished following repeated presentations of the tone without the corneal airpuff.

This critical observation led Martin and Levey to postulate that two parallel conditioning systems might operate simultaneously within the human organism: an overt, somatic-motor preparatory system that requires strict temporal contiguity and statistical contingency to protect the body, and an underlying, covert affective-evaluative system that records the hedonic value of environmental stimuli. Recognizing that their mechanical eyelid apparatus was too narrow to isolate this secondary affective system, they sought to design an entirely new paradigm. They needed an experimental environment stripped of noxious corneal airpuffs, motor apparatuses, and overt somatic demands. By replacing physical reflexes with purely visual aesthetic appraisals, they aimed to isolate the cognitive-affective machinery of associative learning in its purest state.

2.3 Key Publications Establishing the Evaluative Paradigm

The empirical debut of this new conceptual framework arrived in their 1975 landmark paper, “Evaluative Conditioning,” published in Behaviour Research and Therapy. In this concise, methodologically pristine article, Martin and Levey unveiled their visual “picture-picture” pairing methodology using art postcards. The paper demonstrated that pairing an individually calibrated neutral visual stimulus with a liked or disliked visual stimulus reliably shifted the neutral stimulus’s hedonic valence in the direction of the paired item. The findings were statistically robust, showing that these shifts occurred rapidly and without the explicit, overt reinforcement required by Hullian and Skinnerian models.

Three years later, they expanded their empirical footprint with a vital follow-up study: “Evaluative Conditioning: A Survey of Some Initial Findings” (Martin & Levey, 1978). This work probed the boundary conditions and operational parameters of the evaluative paradigm. Here, they systematically manipulated reinforcement schedules, visual presentation intervals, and extinction protocols. The 1978 paper documented the most contentious and enduring characteristic of evaluative conditioning: its apparent resistance to classical extinction. While standard Pavlovian reflexes extinguish rapidly when the CS is repeatedly presented without the US, Martin and Levey demonstrated that subjective hedonic shifts remained remarkably durable, enduring through dozens of non-reinforced presentations with minimal decay.

The culmination of their theoretical work appeared in their 1987 synthesis, “Learning What Will Happen Next: Conditioning and Thinking,” accompanied by their seminal theoretical paper, “Evaluative Conditioning: A Form of Associative Learning with Distinct Properties.” In these later publications, Martin and Levey formally codified the structural dichotomy between classical “expectancy” learning and evaluative “evaluative” learning. They defended their paradigm against mounting cognitive critiques that sought to reduce all associative learning to conscious, declarative hypothesis testing. Through rigorous experimental defense spanning three decades, Martin and Levey firmly established evaluative conditioning as a distinct, irreducible pillar of modern behavioral and cognitive science.

3. The Landmark 1975 Paradigm: Experimental Architecture and Methodology

3.1 Stimulus Selection and Pre-Experimental Standardization

The foundational insight of Martin and Levey’s 1975 experimental architecture was the recognition that hedonic valence is intrinsically subjective; what is aesthetically captivating to one individual may be deeply unappealing or utterly indifferent to another. Consequently, rather than imposing arbitrary, normative stimulus ratings, Martin and Levey instituted an individualized pre-experimental standardization procedure. The visual stimulus library comprised a curated set of 50 photographic reproductions of fine art postcards, encompassing a broad range of artistic traditions, including impressionism, post-impressionism, abstract expressionism, and classical realism. The collection featured paintings by artists such as Paul Cézanne, Claude Monet, Vincent van Gogh, Wassily Kandinsky, and Pablo Picasso, deliberately selected to evoke varied aesthetic and emotional responses.

Upon arrival at the Maudsley laboratory, each participant underwent an idiosyncratic baseline sorting protocol. The participant was seated at a standardized viewing table and presented with the shuffled deck of 50 art reproductions. Their initial task was to execute a tripartite forced-choice sort, categorizing each postcard into one of three physical receptacles designated as:

• Liked (+): Images that evoked an immediate, spontaneous sense of aesthetic pleasure, beauty, or positive emotional attraction.
• Disliked (-): Images that evoked an immediate sense of aesthetic displeasure, revulsion, tension, or negative emotional aversion.
• Neutral (0): Images that evoked complete aesthetic indifference, emotional neutrality, or a distinct lack of qualitative evaluation.

This subjective baseline assessment was crucial. By forcing participants to identify stimuli that sat precisely at their personal hedonic midpoint, Martin and Levey eliminated the noise of population-level normative assumptions. Furthermore, the use of fine art postcards largely controlled for semantic and high-familiarity confounds. Unlike familiar consumer logos or iconic cultural emblems, these varied art postcards lacked pre-existing, overlearned verbal labels, allowing the researchers to measure pristine associative acquisition.

3.2 The Classical Picture-Picture Pairing Procedure

Once the tripartite baseline sort was completed, the experimenter rapidly processed the individual participant’s distributions to construct the conditioning arrays. From the subset of postcards categorized as strictly “Neutral,” the experimenter selected target items to serve as Conditioned Stimuli (CSs). From the “Liked” and “Disliked” categories, items displaying unambiguous affective potency were chosen to serve as positive Unconditioned Stimuli (US+) and negative Unconditioned Stimuli (US-), respectively. Additionally, neutral items were designated to serve as control stimuli (CS0) to be paired with other neutral items (US0), establishing a crucial baseline for spontaneous hedonic drift.

The core conditioning protocol paired these visual stimuli in a series of discrete, contiguous trials. The experimental design featured three primary pairing configurations:

• Positive Pairing (CS+ / US+): A neutral image was systematically paired with a liked art image.
• Negative Pairing (CS- / US-): A neutral image was systematically paired with a disliked art image.
• Neutral Control Pairing (CS0 / US0): A neutral image was systematically paired with another neutral art image.

The physical presentation of these stimulus pairs was achieved using precision slide-projection tachistoscopes, which projected the fine art images onto a rear-projection translucent screen positioned directly in the participant’s visual field. The experimental chamber was light-attenuated and acoustically dampened to prevent external environmental distractions from disrupting associative binding.

3.3 Temporal Dynamics: Inter-Stimulus Intervals and Trial Configurations

In standard Pavlovian conditioning, associative learning is exquisitely sensitive to the Inter-Stimulus Interval (ISI). For autonomic or skeletal reflexes, an ISI of 500 milliseconds is typically optimal; intervals extending beyond several seconds frequently prevent CR acquisition entirely, while simultaneous presentation (ISI = 0) and backward presentation (US preceding CS) typically yield negligible excitatory conditioning. Martin and Levey’s 1975 paradigm, however, operated under radically divergent temporal assumptions. They recognized that visual aesthetic stimuli require sufficient exposure time to permit sensory registration and hedonic appreciation.

In their primary configurations, Martin and Levey implemented both simultaneous presentations (where CS and US images appeared concurrently side-by-side) and very short-delay schedules (where the CS appeared on screen, followed shortly by the contiguous onset of the US, overlapping before simultaneous termination). The typical presentation exposure duration was standardized between 3.0 to 5.0 seconds per trial. This duration was calibrated to allow participants to fully process the artistic composition and emotional resonance of the unconditioned art piece without allowing sufficient time for deliberate cognitive reappraisal or visual scanning boredom.

Crucially, robust evaluative conditioning did not require hundreds of repetitions. While motor reflex conditioning often necessitates extended training sequences, Martin and Levey observed substantial evaluative shifts after remarkably few pairing trials—frequently between 2 to 8 pairings per stimulus set. To minimize sensory habituation and perceptual fatigue, the trial sequence was pseudo-randomized. CS+/US+ pairings, CS-/US- pairings, and CS0/US0 pairings were systematically intermixed with filler trials. Visual masking protocols and brief inter-trial intervals (ITIs) ranging from 10 to 20 seconds of blank, neutral luminance prevented visual after-images and stabilized cortical processing between trials.

3.4 Post-Conditioning Evaluative Assessment Protocols

Following the completion of the conditioning phase, Martin and Levey administered a post-experimental assessment to quantify shifts in subjective valence. The experimental protocol required participants to re-sort the entire library of 50 art postcards using the identical tripartite sorting criteria administered during baseline. The critical analytical question was whether the initially neutral CS items would remain in the neutral receptacle or migrate systematically into the liked or disliked receptacles as a function of their paired US.

To capture finer gradations of evaluative change beyond the coarse tripartite categorical sort, Martin and Levey incorporated refined ordinal and continuous rating protocols. Participants were instructed to sub-divide the broad piles into ordered rankings or to evaluate each postcard along continuous visual analogue scales and Likert-style metrics. These scales ranged from extreme aesthetic displeasure to extreme aesthetic pleasure. The testing environment was rigorously controlled; post-conditioning assessments were conducted using double-blind procedures whenever possible, ensuring that the experimenter administering the sorting deck was unaware of the specific pairing history of the individual cards, thereby eliminating experimenter bias.

The fundamental metric derived from this methodology was the valence delta score ($\Delta V$). Calculated mathematically, the delta score represented the post-conditioning evaluative rating ($R_{\text{post}}$) minus the pre-conditioning baseline rating ($R_{\text{pre}}$) for each conditioned stimulus:
$$\Delta V = R_{\text{post}} – R_{\text{pre}}$$
Martin and Levey’s statistical analyses consistently revealed significant, asymmetrical directional shifts: CS+ stimuli paired with liked paintings exhibited robust, positive delta scores ($\Delta V > 0$), migrating into the “Liked” distribution, whereas CS- stimuli paired with disliked paintings exhibited negative delta scores ($Delta V < 0$), migrating into the "Disliked" distribution. The neutral control stimuli (CS0), conversely, remained stably anchored around the zero-change mark.

4. Stimulus Selection and Rating Systems in the Original Studies

4.1 The Tripartite Sorting Methodology

The operational engine of Martin and Levey’s early studies was their tripartite card-sorting methodology. While modern affective science relies heavily on millisecond-level computerized tracking and implicit reaction-time paradigms, the physical card-sorting task used in 1975 offered distinct psychometric advantages. When presented with physical photographic art reproductions, participants engaged in a tactile, highly focused appraisal process. The mechanical forced-choice distribution forced the participant to actively weigh their aesthetic reactions, converting an amorphous internal feeling state into an overt behavioral placement.

The mechanics of the sorting protocol were executed with geometric precision. The 50 cards, measuring standard postcard dimensions (approximately 10 $\times$ 15 centimeters) and displaying high-fidelity chromatic or monochromatic prints, were thoroughly shuffled. Participants were instructed not to engage in deep intellectual, historical, or technical analysis of the paintings, but rather to rely on immediate, visceral, “gut” aesthetic appraisal. If an image evoked a positive emotional resonance, it was laid to the right; if it provoked aesthetic aversion, boredom, or dissonance, it was laid to the left; if it elicited no emotional stir whatsoever, it was laid in the center.

Following this primary triage, Martin and Levey frequently instructed participants to further sub-divide each of the three primary piles into secondary hierarchies, creating five-point, seven-point, or even nine-point ordinal scales. For instance, the “Liked” pile was systematically sorted into “mildly liked,” “moderately liked,” and “strongly liked.” This multi-tiered psychophysical calibration yielded granular ordinal data, allowing the researchers to identify mathematical thresholds for absolute hedonic neutrality versus marginal or latent preference. Psychometric reliability checks confirmed that human aesthetic preferences for visual art exhibit high internal consistency and test-retest reliability across short retention intervals, providing a stable baseline against which associative shifts could be measured.

4.2 Hedonic Calibration and Subjective Neutrality

One of the most profound theoretical and methodological hurdles confronting evaluative conditioning research is the calibration of true subjective neutrality. In a strict psychophysical sense, what constitutes a “neutral” stimulus? A stimulus categorized as neutral might represent a true hedonic zero—a point of complete emotional indifference. Alternatively, it might represent a point of profound ambivalence, wherein high positive affective elements and high negative affective elements are simultaneously experienced, canceling one another out into a tense, unstable equilibrium. A third possibility is that a neutral placement merely reflects cognitive apathy or perceptual boredom.

Martin and Levey recognized that if a stimulus was not a true hedonic zero, but rather an unstable ambivalent point, it would be vulnerable to spontaneous baseline shifts in the absence of associative learning. To insulate their experimental design against regression to the mean and baseline instability, they established strict selection criteria for both CS and US items. A postcard was certified as a viable neutral CS only if it sat comfortably in the central cluster during both primary and secondary sorts, showing zero drift during pre-experimental calibration phases.

Conversely, the selection of Unconditioned Stimuli required unambiguous, polar affective potency. A liked painting had to sit definitively in the upper quartile of the positive distribution, while a disliked painting had to anchor the lowest quartile of the negative distribution. Martin and Levey observed wide individual differences in aesthetic reactivity. Some participants displayed broad, bell-shaped affective distributions, generating expansive neutral piles and small, highly selective hedonic extremes. Other participants displayed bi-modal, polarized distributions, finding nearly every artistic work either deeply compelling or intensely irritating. By personalizing stimulus selection based on each participant’s baseline distribution spread, Martin and Levey guaranteed that the affective distance between the CS and the US was maximized on an individualized, psychometrically calibrated basis.

4.3 The Visual Stimulus Library: Art Reproductions as Affective Tools

The selection of fine art postcards as the primary stimulus library was a deliberate, sophisticated methodological choice. Orthodox conditioning paradigms frequently employed simple geometric forms, monochromatic color swatches, or synthetic linguistic tokens (nonsense syllables or consonant trigrams). While these simple stimuli offered ease of experimental manipulation, Martin and Levey argued that they suffered from profound structural limitations in the study of complex human affect.

Geometric shapes and raw colors lack the evocative visual complexity necessary to elicit rich, authentic hedonic reactions in human adults. When presented with a blue triangle or a yellow circle, human participants rarely experience genuine aesthetic joy or visceral revulsion; their reactions are flat, cognitive, and artificial. Conversely, linguistic tokens such as words or nonsense syllables immediately trigger pre-existing semantic, syntactic, and autobiographical networks, introducing unpredictable cognitive interference. Complex fine art reproductions occupied a unique methodological middle ground. They provided rich, multidimensional visual compositions capable of generating deep, genuine emotional and aesthetic reactions, while remaining relatively free from intrusive pre-existing verbal associations or episodic memories.

Martin and Levey carefully balanced the structural properties of their art library. The collection contained a strategic mixture of abstract expressionist paintings (e.g., works by Jackson Pollock or Mark Rothko) and photographic realism or figurative compositions (e.g., portraits by Rembrandt or landscapes by John Constable). Abstract works were exceptionally useful because they bypassed semantic categorizations; their aesthetic impact relied entirely on color harmonics, balance, line turbulence, and raw perceptual composition. Figurative works, while introducing some thematic content, provided reliable access to established hedonic valences. Subsequent empirical evaluations supported Martin and Levey’s intuition: complex artistic visual stimuli produced significantly larger and more robust evaluative conditioning effect sizes than simplistic, synthetic shapes, establishing the art-postcard paradigm as an enduring gold standard in early evaluative research.

5. Distinguishing Evaluative Conditioning from Classical Pavlovian Conditioning

5.1 Signal Value versus Value Transfer: CS-US Relationships

The most consequential theoretical contribution of Martin and Levey’s work was their insistence that evaluative conditioning is not merely a cosmetic variant of classical Pavlovian conditioning, but an associative phenomenon operating through an entirely different functional architecture. The conceptual core of this distinction lies in the divergence between signal value and value transfer.

In classical Pavlovian conditioning, the Conditioned Stimulus operates primarily as a predictive signal. Through repeated, temporally contiguous pairings, the CS informs the organism about an impending environmental event: “Stimulus A predicts the arrival of Stimulus B.” The animal or human does not confuse the CS with the US; a dog does not attempt to chew or digest the auditory bell that signals meat powder. Rather, the bell acquires signal value, prompting the animal to execute an anticipatory, preparatory response (salivation) tailored to the upcoming unconditioned event. The associative relationship is referential, prospective, and rooted in environmental utility. This aligns neatly with Edward Tolman’s cognitive mapping framework and modern expectancy models: the organism builds an internal predictive model of environmental states.

In evaluative conditioning, as conceptualized by Martin and Levey, the CS acquires no predictive signal value whatsoever. The participant does not view the neutral art postcard as a herald warning that a disliked postcard is about to appear. Instead, evaluative conditioning represents direct hedonic value transfer: the neutral CS directly acquires the intrinsic emotional and aesthetic properties of the US. The CS becomes “infused” with the valence of the item with which it co-occurred. After pairing with a liked painting, the neutral painting does not mean “a nice painting is coming soon”; rather, the neutral painting is now perceived as “a nice painting in and of itself.” There is no preparatory motor response, no somatic readiness, and no behavioral anticipation. Evaluative conditioning alters the qualitative perception of the object itself, representing a structural transformation of affective value rather than the calculation of a predictive environmental contingency.

5.2 Sensitivity to Statistical Contingency and Presentation Schedules

Classical Pavlovian conditioning is governed by statistical contingency. As demonstrated mathematically by Robert Rescorla in the late 1960s, temporal contiguity alone is insufficient to produce robust classical conditioning. An animal will only acquire a conditioned response if the CS provides predictive information about the US—that is, if the conditional probability of the US given the presence of the CS, $P(\text{US}|\text{CS})$, is significantly higher than the probability of the US in the absence of the CS, $P(\text{US}|\neg\text{CS})$. If a US is delivered randomly in the background, contingency is degraded, and classical conditioning fails to emerge. Furthermore, classical conditioning is profoundly susceptible to phenomena such as blocking (where prior conditioning to one cue blocks learning about a redundant second cue) and overshadowing.

Evaluative conditioning defies these classical statistical constraints in striking ways. In their 1978 and 1987 investigations, Martin and Levey, alongside contemporary European researchers, discovered that evaluative conditioning is remarkably insensitive to degraded statistical contingency. If an experimenter implements a partial reinforcement schedule—presenting the CS multiple times in isolation alongside occasional CS-US pairings—classical reflex conditioning declines dramatically. Evaluative shifts, however, remain largely intact. The hedonic transfer that occurs during the contiguous pairing trials appears insulated from the non-reinforcement trials that surround it.

Similarly, early evaluative conditioning paradigms consistently failed to demonstrate classical blocking and overshadowing effects. In an evaluative context, if two neutral stimuli are simultaneously paired with an aversive US, both stimuli tend to acquire negative valence, regardless of whether one stimulus was previously conditioned or perceptually dominant. Evaluative learning does not behave like an information-processing system computing the statistical necessity of an environmental predictor. Instead, it behaves like an automatic, spatiotemporal binding mechanism: whenever a neutral stimulus enters the perceptual and affective orbit of a valenced stimulus, hedonic transfer occurs, relatively unconstrained by the probabilistic mathematical contingencies that govern classical reflex learning.

5.3 Temporal Symmetry and Directional Conditioning

Perhaps the most radical empirical divergence between classical reflexology and Martin and Levey’s evaluative model lies in the domain of temporal symmetry. Classical Pavlovian learning exhibits strict temporal directionality. For an organism to establish an adaptive preparatory response, the CS must precede the US. Optimal conditioning occurs in forward delay conditioning schedules, where the onset of the CS occurs shortly before the onset of the US. If the CS and US are presented simultaneously (ISI = 0), classical conditioning is notoriously weak or non-existent, because a simultaneous cue provides zero predictive utility; the unconditioned event is already occurring, rendering a preparatory signal functionally useless. If a backward conditioning schedule is implemented (US preceding the CS), classical excitatory conditioning virtually never develops, often producing inhibitory conditioning instead.

Martin and Levey demonstrated that evaluative conditioning exhibits profound temporal symmetry. In their 1975 and subsequent experiments, simultaneous presentation of the CS and US fine art images produced evaluative shifts that were fully equivalent, and often superior, to forward delay presentations. Because evaluative conditioning is an integrative, holistic binding process rather than an anticipatory warning system, the simultaneous perceptual co-occurrence of the two images provides the ideal conditions for affective integration. The cognitive visual system captures both items within the same temporal window, allowing the emotional resonance of the US to color the perceptual processing of the CS.

Furthermore, multiple researchers working within the Martin-Levey paradigm documented significant backward evaluative conditioning: presenting a highly disliked or liked art image immediately *before* presenting a neutral image also resulted in reliable valence transfer to the neutral item. While backward conditioning is an anomaly or an impossibility within orthodox predictive models, it fits seamlessly within Martin and Levey’s framework. If the unconditioned stimulus activates an internal affective or emotional state that outlasts the physical offset of the image, any neutral stimulus presented immediately into that lingering emotional wake will become associated with that qualitative state. Evaluative conditioning does not care which stimulus arrived first; it only requires that the internal affective representation of the US co-exists temporally with the perceptual representation of the CS.

6. Extinction Resistance and the Durability of Evaluative Shifts

6.1 The Phenomenon of Non-Extinction in Evaluative Conditioning

In the canon of classical conditioning, the law of extinction is absolute. If a conditioned stimulus is repeatedly and reliably presented in the complete absence of the unconditioned stimulus, the conditioned response steadily decays, eventually returning to baseline levels. A dog that hears an auditory bell without receiving meat powder will progressively stop salivating; a human subject exposed to an auditory tone without the accompanying corneal airpuff will extinguish their conditioned eyeblink response within dozens of trials. Within Pavlovian and Hullian theory, extinction does not erase the original associative trace, but rather establishes an active, inhibitory learning layer: the organism learns the new contingency that “the CS now predicts the absence of the US.”

One of Martin and Levey’s most startling and controversial empirical discoveries was the apparent non-extinction of evaluative conditioning. In their 1978 and 1987 publications, they detailed experiments wherein CS+ and CS- visual stimuli, having acquired liked and disliked valences during the initial pairing phase, were repeatedly presented to participants completely alone, without the accompanying US images. In traditional conditioning, these repeated non-reinforced exposures would rapidly drive the delta scores back to zero. In Martin and Levey’s evaluative paradigms, the acquired valence shifts remained virtually indestructible.

Participants continued to categorize the CS+ paintings as liked and the CS- paintings as disliked through dozens of non-reinforced extinction presentations. The acquired hedonic shift demonstrated an extraordinary durability that defied orthodox associative expectations. Martin and Levey postulated that evaluative learning represents an indelible memorial alteration. Because the evaluative response is not an expectancy of an external event, the mere presence of the CS alone does not generate an “expectancy violation.” In classical conditioning, the animal expects food, receives nothing, notes the error, and adjusts its predictive behavior. In evaluative conditioning, the subject looks at the CS and feels aesthetic displeasure. Because there was never an expectation of a secondary event occurring, the experience of looking at the card and feeling displeasure simply confirms its negative status. The affective evaluation reinforces itself, rendering the conditioned hedonic valence fundamentally resistant to standard extinction protocols.

6.2 US Revaluation and Counterconditioning Paradigms

The unique structural durability of evaluative conditioning is further illustrated by its complex interactions with US revaluation and counterconditioning paradigms. In classical conditioning, modifying the psychological value of the US after the conditioning phase has concluded—a process known as post-conditioning US revaluation—frequently alters the magnitude of the conditioned response. For instance, if an animal is conditioned to associate a tone with food, and the food is subsequently paired with lithium chloride to induce nausea while the tone is never played, subsequent presentation of the tone alone typically evokes diminished salivation. The animal updates its internal representation of the food, and the predictive CS reflects this updated value.

In early evaluative conditioning research conducted by Martin, Levey, and subsequent investigators such as Baeyens and Eelen, the effects of US revaluation were often elusive and highly constrained. When researchers paired a neutral CS with a liked US, and subsequently degraded the value of that US (for instance, by presenting it repeatedly until participants experienced profound aesthetic boredom or satiety), the previously established evaluative shift on the CS frequently persisted unperturbed. Once the hedonic quality had successfully migrated to the CS during the initial pairing, the CS appeared to achieve affective autonomy. It no longer relied upon an active, real-time computational link back to the originating US representation; it had become an inherently valenced aesthetic object in its own right.

This autonomy is equally evident in counterconditioning paradigms. If an experimenter attempts to reverse a previously acquired evaluative shift—taking an established CS- that has been paired with a disliked image and subsequently pairing it with an intensely liked US+—the outcome differs markedly from classical systems. Rather than smoothly canceling out or overwriting the original negative valence, the stimulus frequently exhibits profound resistance, requiring an inordinate number of positive pairings to shift its hedonic baseline. In many cases, participants exhibit a form of evaluative ambivalence or affective dissonance: the acquired negative valence forms an enduring, indelible substrate in associative memory, demonstrating that while evaluative conditioning is remarkably easy to establish, it is exceptionally difficult to dismantle.

6.3 Long-Term Retention and Temporal Decay Profiles

To establish the external validity and theoretical depth of evaluative conditioning, researchers moved beyond immediate post-experimental testing to examine long-term retention and temporal decay profiles. Standard laboratory-acquired physiological reflexes, when not periodically refreshed by reinforcement, are notorious for degrading over extended retention intervals. Memory decay, spontaneous recovery, contextual drift, and retroactive interference rapidly degrade conditioned autonomic and motor responses over days and weeks.

Martin and Levey, followed by researchers such as Dirk Baeyens, Frank Eelen, and later modern social psychologists, tested participants across delayed post-testing intervals ranging from 24 hours to several months post-exposure. The empirical results consistently demonstrated that evaluative shifts exhibit an unprecedented level of temporal persistence. Participants who were brought back to the laboratory weeks after the initial picture-picture pairing phase continued to demonstrate significant, directional valence delta scores for the conditioned art postcards. Even when participants could no longer consciously recall which art postcard had been paired with which, their subjective liking and disliking metrics remained tilted in the conditioned directions.

This remarkable durability points toward specialized memory consolidation mechanisms underlying evaluative acquisition. Evaluative shifts appear to integrate directly into implicit, semantic, and affective memory networks. Unlike declarative memories of episodic events, which are subject to rapid synaptic decay and hippocampal reorganization, acquired hedonic valences appear to be stored as stable, non-declarative affective tags. These enduring affective modifications operate beneath the threshold of explicit episodic recall, providing organisms with a persistent, low-cost hedonic library of environmental objects that continues to bias preference and avoidance behavior long after the initial contextual details of the pairing event have faded into oblivion.

7. The Role of Contingency Awareness in Martin and Levey’s Framework

7.1 Defining and Measuring Awareness in Early Studies

No issue in the history of evaluative conditioning has generated more fierce debate, methodological scrutiny, and theoretical controversy than the role of contingency awareness. Within associative learning theory, contingency awareness is defined as the subject’s explicit, conscious, declarative knowledge of the experimental contingencies—specifically, the conscious awareness that Stimulus A was systematically paired with Stimulus B during the experimental presentation phase.

In their seminal 1975 study and their 1978 follow-up investigations, Martin and Levey operated under the hypothesis that evaluative conditioning could occur entirely without declarative contingency awareness. To evaluate this claim, they instituted post-experimental retrospective questionnaires. After completing all conditioning trials and the final post-experimental card sorts, participants were subjected to structured debriefing interviews. The experimenter asked open-ended questions designed to determine whether the participant had deduced the associative pairings:

• “Did you notice any connection or relationship between any of the cards presented on the screen?”
• “Did certain neutral cards consistently appear with specific pleasant or unpleasant cards?”
• “What do you believe was the purpose of the slide presentations?”

Based on these retrospective verbal reports, Martin and Levey classified participants into “aware” and “unaware” subsets. Their analyses revealed an extraordinary finding: significant, directional evaluative conditioning shifts occurred in participants who demonstrated zero declarative recall of the CS-US pairings. Subjects who were completely unable to identify which neutral card had been paired with which liked or disliked card nonetheless exhibited robust valence delta scores, sorting the CS+ items into the liked receptacle and the CS- items into the disliked receptacle. Martin and Levey concluded that subjective preference shifts were driven by an automatic, implicit associative mechanism that did not depend on conscious, declarative contingency calculation.

7.2 The Implicit-Explicit Dichotomy in Evaluative Learning

Martin and Levey’s conclusion positioned evaluative conditioning at the forefront of the cognitive revolution’s exploration of implicit versus explicit psychological systems. By arguing that evaluative learning could operate independently of contingency awareness, they conceptualized evaluative conditioning as a distinct, non-declarative learning module. This perspective aligned their work with contemporaneous breakthroughs in neuropsychology, such as the discovery that amnesic patients with severe bilateral hippocampal damage (such as patient H.M.) could acquire motor skills, perceptual priming, and certain affective preferences despite possessing zero conscious, declarative memory of the learning episodes.

This implicit-explicit dichotomy forged deep theoretical connections between evaluative conditioning and other non-declarative affective phenomena, most notably the Mere Exposure Effect pioneered by Robert Zajonc. In 1980, Zajonc published his landmark paper, “Preferences Need No Inferences,” arguing that affective processing occurs faster than, and independently of, conscious cognitive appraisals. Zajonc demonstrated that subliminal, unremembered exposure to novel stimuli increases subsequent positive liking. Martin and Levey’s work provided the associative counterpart to Zajonc’s mere exposure paradigm: while Zajonc demonstrated that unremembered exposure breeds familiarity and liking, Martin and Levey demonstrated that unremembered contiguous pairing breeds the bidirectional transfer of both positive and negative hedonic valence.

Within this theoretical framework, the human affective architecture was viewed as a dual-layered system. One layer consists of explicit, declarative, propositional cognition: it consciously tracks environmental probabilities, calculates risks, forms verbal hypotheses, and guides intentional motor planning. The second, deeper layer consists of an implicit, non-declarative associative network: it silently and continuously tracks spatiotemporal coincidences, binding hedonic valences to perceptual representations without requiring the spotlight of conscious attention. Evaluative conditioning was crowned as the archetypal empirical manifestation of this implicit affective system.

7.3 The Awareness Controversy: Epiphenomenon or Prerequisite?

The claim that evaluative conditioning operates completely in the absence of conscious contingency awareness did not stand unchallenged. Beginning in the late 1980s and intensifying through the 1990s and 2000s, cognitive learning theorists mounted a formidable critique against Martin and Levey’s awareness methodology. Leading this counter-offensive were cognitive psychologists such as David R. Shanks and Peter F. Lovibond, who argued that Martin and Levey’s early awareness checks suffered from profound measurement insensitivity.

Shanks and Lovibond pointed out that retrospective questionnaires administered at the conclusion of an experiment are fundamentally inadequate tools for capturing fleeting conscious awareness that may have been present during the actual learning trials. A participant might be fully conscious of the pairing on Trial 3, experience an immediate hedonic shift, and subsequently forget the specific pairing episode by the time the post-experimental interview is conducted 20 minutes later. In this scenario, the verbal report reveals a failure of retrospective memory retrieval, not the operation of an unconscious conditioning process. Furthermore, critics argued that free-recall verbal probes lack sensitivity compared to forced-choice recognition tests; participants who claim complete ignorance on an open-ended interview frequently perform well above chance when forced to select the correct CS-US pair from an array.

This methodological dispute sparked the great “Awareness Controversy.” One intellectual camp, championing the cognitive-propositional model, asserted that conscious contingency awareness is an indispensable prerequisite for all human associative learning, including evaluative shifts. They argued that evaluative conditioning only occurs when participants explicitly recognize and construct propositions about the CS-US co-occurrence (e.g., “This postcard is presented alongside that grotesque abstract painting”). The opposing camp, defending and refining Martin and Levey’s original thesis, argued that while contingency awareness can certainly enhance evaluative acquisition through top-down cognitive channels, it is not a structural prerequisite. Over subsequent decades, this controversy stimulated remarkable methodological innovations, including trial-by-trial concurrent awareness probes, gaze-tracking, secondary cognitive-load manipulations, and subliminal stimulus presentation, solidifying the Martin-Levey awareness debate as one of the most intellectually fruitful battlegrounds in cognitive psychology.

8. Cognitive and Affective Processing Mechanisms in Early EC Models

8.1 Holistic versus Analytical Processing Models

To provide a rigorous theoretical account for why evaluative conditioning exhibits properties so strikingly different from classical reflex conditioning, Martin and Levey formulated an innovative model of visual and affective information processing. They contrasted two distinct modes of human perceptual integration: analytical processing and holistic processing.

Analytical processing is the hallmark of traditional classical conditioning and predictive contingency detection. In an analytical mode, the human cognitive apparatus actively decomposes a complex visual or sensory environment into discrete, individual features. The organism isolates the conditioned stimulus from the surrounding context, monitors its precise temporal boundaries, and assesses its predictive validity relative to the unconditioned stimulus. This analytical mode is computationally expensive, highly reliant upon focal attention, and exquisitely tuned to statistical contingencies, logical contradictions, and temporal directionality. If an analytical processor observes that a CS does not reliably predict a US, it dismantles the associative expectancy.

In contrast, Martin and Levey postulated that evaluative conditioning is primarily governed by holistic processing. When a participant is exposed to the simultaneous or near-simultaneous presentation of two complex fine art postcards, the visual system does not necessarily engage in rigid, feature-by-feature temporal analysis. Instead, the perceptual and emotional features of both images are grasped simultaneously within a single, integrated Gestalt field. The emotional resonance, color harmony, visual mood, and hedonic tone of the unconditioned painting wash over the entire visual perceptual window. Within this holistic representation, the neutral CS is not processed as an isolated predictive signal, but as an embedded component of an integrated, affective visual episode. The hedonic quality is infused into the perceptual representation of the CS precisely because, at the moment of affective processing, the two items form an indissoluble perceptual whole.

8.2 Referential versus Expectancy Learning Mechanics

The theoretical insights of Martin and Levey were significantly refined and formalized through their close collaboration and intellectual dialogue with European learning theorists, most notably Dirk Baeyens and Frank Eelen at the University of Leuven. This collaborative evolution crystallized into the definitive theoretical distinction between expectancy learning and referential learning.

Expectancy learning is the computational engine of standard classical Pavlovian conditioning. Within an expectancy architecture, the formation of an associative link between CS and US leads the presentation of the CS to activate an explicit mental expectancy of the impending physical arrival of the US:
$$\text{CS} long\rightarrow \text{Expectancy of US Arrival} long\rightarrow \text{Preparatory CR}$$
Because this system generates a predictive expectation, it is subject to immediate revision if the US fails to materialize. It requires forward temporal contiguity, relies heavily on cognitive processing resources, and extinguishes rapidly when environmental conditions change.

Evaluative conditioning, Martin and Levey argued, operates via a referential learning architecture. In a referential learning system, the associative link between the CS and US does not generate an expectancy of an upcoming event. Rather, the presentation of the CS merely activates the internal, stored mental representation of the US:
$$\text{CS} long\rightarrow \text{Activation of US Memory Representation} long\rightarrow \text{Direct Affective Resonances}$$
When the neutral art postcard is subsequently presented during the post-test, it does not lead the subject to expect the arrival of the liked painting. Instead, the CS simply calls to mind, at an automatic or semi-automatic level, the aesthetic, semantic, and hedonic features of the paired item. Because the referential link does not predict a future occurrence, the non-occurrence of the US does not produce an expectancy violation. There is no predictive error signal to drive extinction learning. The referential link remains structurally stable, quietly routing the perceptual input of the CS through the qualitative hedonic circuits originally established by the US.

8.3 Affect-as-Information and Semantic Feature Transfer

In exploring the cognitive mechanics of this value transfer, early evaluative models intersected with emergent theories of affective judgment, particularly the Affect-as-Information hypothesis formulated by Gerald Clore and Norbert Schwarz. This framework posits that when human beings are confronted with an object and asked to make an evaluative judgment (e.g., “Do I like this painting?”), they frequently use their immediate internal feeling state as a diagnostic source of information: “How do I feel about it?”

When a participant views a CS postcard that has undergone evaluative conditioning, the presentation of the image activates the referential associative trace. This activation induces a micro-affective reaction—a subtle internal feeling of pleasantness or aversion. Rather than attributing this internal feeling to the historical pairing phase, the human cognitive system automatically attributes the affective feeling to the object itself. The subject misattributes the lingering affective echo of the US as an intrinsic, structural quality of the CS: “I feel an internal sensation of tension and dissonance when looking at this postcard; therefore, this postcard must be an aesthetically unpleasant work of art.”

Importantly, research expanding from Martin and Levey’s paradigm demonstrated that evaluative conditioning is not limited to the transfer of crude, undifferentiated hedonic valence (good versus bad). It also facilitates the transfer of specific, qualitative semantic and emotional features. If a neutral visual stimulus is paired with a painting that evokes quiet, pastoral tranquility, the CS acquires a tranquil, peaceful valence. If the same neutral stimulus is paired with a painting that evokes chaotic, aggressive, violent horror, it acquires an aggressive, agitated negative valence. Evaluative conditioning does not merely assign a mathematical plus or minus sign to an internal node; it sculpts the specific qualitative affective texture of internal representations, seamlessly blending associative learning mechanisms with complex human cognitive appraisal systems.

9. Methodological Critiques, Replication Attempts, and Conceptual Controversies

9.1 Demand Characteristics and Experimenter Bias Critiques

Almost immediately following the publication of Martin and Levey’s 1975 and 1978 findings, critics challenged the paradigm on methodological grounds, asserting that the reported valence shifts were the product of demand characteristics and subtle experimenter bias. Leading critics, such as Hollands and colleagues, suggested that human participants placed in an artificial laboratory environment are exceptionally adept at deducing the underlying research hypothesis. If a subject recognizes that the experimenter has repeatedly presented a specific neutral painting alongside a universally adored Monet landscape, they may infer that the “correct,” socially compliant behavior during the post-test is to assign that neutral painting a positive rating.

The forced-choice tripartite sorting methodology was singled out as a primary vector for such inflation. Critics argued that forcing participants to re-sort the cards into positive, negative, and neutral receptacles artificially amplified microscopic or non-existent preferences. In the absence of an explicit experimental pairing, a participant might leave a card in the neutral pile indefinitely. But if forced to make subtle distinctions, any faint, conscious memory of having seen that card displayed alongside a pleasant painting would lead the subject to consciously categorize it as “Liked” simply to satisfy the perceived demands of the laboratory protocol.

Martin and Levey vigorously defended their paradigm against these critiques through extensive empirical counter-measures. They introduced complex visual filler items, deceptive instructional framing, and sophisticated post-experimental interviews specifically calibrated to detect demand compliance. In several experiments, they utilized a double-blind protocol where the research assistants conducting the post-conditioning card sorting tests were entirely unaware of which cards had served as CS+, CS-, or control stimuli for that specific participant. Even under these rigorous, double-blind controls, where subtle experimenter cuing was physically impossible, the directional valence delta scores remained statistically significant. While demand characteristics could undeniably contaminate poorly controlled associative studies, Martin and Levey demonstrated that genuine evaluative shifts occurred under conditions where demand explanations were thoroughly controlled.

9.2 Replication Difficulties: The Hollands, Rozin, and Field Challenges

Despite Martin and Levey’s rigorous defenses, the broader scientific community encountered notable difficulties in replicating consistent evaluative conditioning effects across disparate laboratory settings throughout the late 1980s and 1990s. One prominent challenge came from Paul Rozin and his colleagues at the University of Pennsylvania. Rozin, a pioneering researcher in the psychology of disgust and taste learning, attempted to translate evaluative conditioning paradigms into the gustatory and olfactory domains. While Rozin successfully documented strong evaluative shifts in conditioned taste aversions (where a novel flavor is paired with nausea), his laboratory frequently failed to replicate Martin and Levey’s visual picture-picture evaluative effects when strict controls against contingency awareness were introduced.

Similarly, a contentious series of replication attempts by Hollands, Rozin, and later meta-analytic scrutiny by Andy Field, cast doubt upon the ubiquity of unconscious evaluative conditioning. Field published several highly critical evaluations asserting that early evaluative conditioning literature was distorted by severe publication bias, small sample sizes, and underpowered statistical designs. Field argued that when modern, heavily powered, computerized replication studies were executed, the evaluative conditioning effect size shrank dramatically, and often disappeared entirely among participants who demonstrated complete lack of contingency awareness.

These replication controversies sparked intense scientific introspection, ultimately revealing that evaluative conditioning is bounded by subtle, crucial boundary conditions. Discrepant findings between laboratories were traced to minute procedural variations:

• Stimulus Exposure Duration: Studies utilizing exposure times that were too brief (under 1 second) failed to allow full aesthetic processing of the US, while exposures that were too long (over 10 seconds) invited bored cognitive counter-arguing.
• Hedonic Distance: Paradigms that utilized weakly valenced or ambiguous US items produced negligible valence transfer compared to Martin and Levey’s deeply polarized fine art library.
• Sensory Modality: Visual-visual pairings operate under holistic Gestalt mechanisms that do not directly translate to cross-modal pairings (e.g., visual CS paired with an auditory US), which engage analytical processing channels more readily.

Rather than invalidating Martin and Levey’s seminal discoveries, these replication challenges deepened the scientific community’s understanding of the precise cognitive and environmental parameters required to reliably produce evaluative learning.

9.3 Measurement Sensitivity and Statistical Controversies

A third front of methodological critique focused on the psychometric and statistical foundations of the early 1975 paradigm. In their foundational work, Martin and Levey utilized ordinal card sorting and coarse Likert categories. Psychometricians pointed out that treating ordinal sorting changes as continuous interval data in parametric statistical analyses (such as Analysis of Variance) was mathematically problematic and susceptible to Type I error inflation.

Furthermore, early statistical designs faced challenges regarding regression toward the mean. If an experimenter selects a stimulus precisely because it was rated at the extreme midpoint (neutrality) during a noisy, variable baseline test, any subsequent re-test will naturally display statistical dispersion toward positive or negative ends simply due to measurement error. Critics suggested that what Martin and Levey observed was not necessarily true associative transfer, but the directional capture of random measurement regression.

In response to these statistical critiques, the evaluative conditioning methodology underwent a major technological and psychometric transformation through the 1980s and 1990s. Researchers abandoned physical cardboard sorts in favor of continuous Visual Analogue Scales (VAS), where participants mark their preferences along an unbroken, 100-millimeter computerized continuum. Advanced statistical models, such as mixed-effects linear modeling and multi-level hierarchical regressions, replaced basic ANOVA models, allowing researchers to systematically control for individual baseline variances and regression toward the mean. When modern quantitative researchers conducted rigorous re-analyses of Martin and Levey’s original foundational datasets using modern statistical techniques, the core findings held firm: the directional interaction between stimulus pairing and valence shift remained statistically robust, confirming the empirical reality of the phenomenon.

10. Subsequent Refinements and Theoretical Evolutions (1980s–1990s)

10.1 Martin and Levey’s 1987 Theoretical Synthesis

Twelve years after their initial publication, Irene Martin and A.B. Levey published their definitive theoretical magnum opus: “Learning What Will Happen Next: Conditioning and Thinking” (1987). In this landmark monograph, they systematically unified their empirical findings into a comprehensive dual-process conditioning framework. They directly confronted the cognitive revisionism that was sweeping through contemporary learning theory—a revisionism that sought to subsume all classical conditioning into conscious human propositional hypothesis testing.

Martin and Levey defended the classical associative roots of learning while simultaneously establishing the autonomous status of evaluative processes. They systematically categorized the profound procedural, operational, and functional divergences separating classical preparatory conditioning from evaluative conditioning. They argued that while cognitive, propositional expectancy learning had evolved to navigate the fluid, ever-changing predictive relationships of the physical environment, evaluative learning represented an older, more primal, and biologically vital evolutionary mechanism.

From an evolutionary perspective, Martin and Levey reasoned that an organism cannot afford to consciously deliberate or recalculate the hedonic value of every encountered environmental object from scratch. Once an association between a novel object and an intrinsically positive (nutritious, socially rewarding, safe) or negative (toxic, violent, diseased) stimulus has been experienced, it is highly adaptive for the organism to permanently “tag” that novel object with a durable hedonic valence. This evolutionary mandate explains the profound extinction resistance, the temporal symmetry, and the contingency insensitivity of evaluative conditioning. Through this 1987 synthesis, Martin and Levey provided the intellectual scaffolding that rescued associative learning from purely mechanical reflexology and positioned it as an indispensable architecture of modern affective cognitive science.

10.2 The Functional-Cognitive Debate: Baeyens, Eelen, and the Louvain Group

The baton of evaluative conditioning research was enthusiastically taken up in the late 1980s and 1990s by the legendary “Louvain School” at the University of Leuven in Belgium, led by Dirk Baeyens, Frank Eelen, and their colleagues (including Geert Crombez and Jan De Houwer). The Louvain group expanded Martin and Levey’s initial findings into one of the most rigorously productive experimental programs in European psychology.

Baeyens and Eelen formalized and empirically verified the referential learning model, conducting hundreds of experiments that methodically tested the boundaries of Martin and Levey’s original assertions. They expanded evaluative conditioning far beyond visual fine art postcards, designing elegant paradigms across multiple sensory modalities:

• Gustatory Conditioning: Pairing neutral, unfamiliar chemical flavors (e.g., novel fruit essences) with strongly sweet (pleasant) or intensely bitter (aversive, e.g., Tween20) taste solutions.
• Olfactory Conditioning: Pairing neutral visual cues with pleasant fragrances (e.g., vanillin) or putrid odors (e.g., hydrogen sulfide).
• Auditory Paradigms: Pairing neutral acoustic frequencies with harmonic musical chords or dissonant acoustic screeches.

The Louvain group also introduced sophisticated concurrent awareness indices, such as real-time trial-by-trial predictive betting, to capture contingency knowledge in real time. Their exhaustive empirical corpus largely vindicated Martin and Levey’s core postulates: evaluative shifts across multiple sensory modalities demonstrated pronounced extinction resistance, high resistance to statistical degradation, and consistent dissociations from conscious expectancy learning, establishing the Louvain School as the spiritual and scientific successor to the Maudsley laboratory.

10.3 Integration into Contemporary Dual-Process Theories

As cognitive science entered the twenty-first century, the insights of Martin and Levey were mapped directly onto emergent dual-process architectures of human cognition, most famously popularized by Daniel Kahneman, Keith Stanovich, and Richard West as System 1 and System 2. Within this conceptual architecture, evaluative conditioning was identified as an archetypal System 1 process: fast, automatic, associative, non-declarative, context-insensitive, and requiring minimal cognitive working memory resources. Classical expectancy conditioning, conversely, aligned with System 2: slow, deliberative, rule-governed, proposition-based, and heavily reliant on conscious attention.

However, this integration ignited a fierce contemporary debate led by cognitive psychologist Jan De Houwer. De Houwer advocated for a critical conceptual distinction between the functional definition of evaluative conditioning and the mechanistic assumptions underlying it. Functionally, De Houwer defined evaluative conditioning purely as an observed effect: a change in the valence of a stimulus resulting from the pairing of stimuli. Mechanistically, however, De Houwer challenged the assumption that this effect is inevitably mediated by non-declarative, low-level associative links. He formulated a comprehensive propositional model of evaluative conditioning.

The propositional model posits that all evaluative shifts—even those that appear fast and automatic—are ultimately mediated by the rapid, micro-level generation and evaluation of cognitive propositions regarding the relationship between the stimuli (e.g., “This object is paired with that positive object, indicating a shared quality”). This functional-cognitive debate represents the direct modern continuation of the intellectual questions first opened by Martin and Levey in 1975: Does associative learning fundamentally rewire the hedonic fabric of human memory via automatic, low-level binding, or is human affect perpetually governed by the swift, silent operation of internal propositional thought?

11. Modern Applications: From Attitude Formation to Clinical Psychology

11.1 Attitude Formation and Consumer Advertising Paradigms

While Martin and Levey conceived evaluative conditioning within an academic psychiatric laboratory, their findings provided the empirical key to understanding the multibillion-dollar engine of modern consumer advertising and brand cultivation. For decades, commercial advertising had operated on an intuitive, unscientific premise: surround a corporate product with beautiful, sexually attractive, or prestigious imagery, and sales will rise. Martin and Levey’s 1975 paradigm provided the formal empirical and theoretical explanation for precisely how and why this strategy succeeds.

In modern consumer psychology, an unfamiliar corporate logo, novel packaging design, or brand name operates precisely as an initially neutral Conditioned Stimulus ($CS_0$). Advertisers systematically pair this neutral CS with intensely valenced Unconditioned Stimuli ($US^+$), such as charismatic celebrities, breathtaking landscapes, athletic triumphs, or scenes of warm familial joy. Applying the Martin-Levey architecture, consumer psychologists demonstrated that this contiguous visual co-occurrence drives a direct, automatic transfer of hedonic valence to the brand logo. Crucially, in modern high-distraction media environments—such as scrolling through digital social media feeds or walking down chaotic supermarket aisles—consumers rarely engage in analytical, System 2 propositional evaluation of products. Instead, they rely on rapid, holistic, heuristic visual appraisals.

Because evaluative conditioning exhibits remarkable extinction resistance, a brand that successfully binds itself to positive hedonic valence maintains that halo effect even when the consumer is removed from the advertising environment. Furthermore, consistent with Martin and Levey’s findings regarding contingency awareness, consumers frequently demonstrate an acquired emotional preference for a conditioned brand even when they cannot consciously remember which commercial or context originally featured the pairing. This non-declarative preference transfer has raised profound ethical questions regarding the pervasive, algorithmic deployment of evaluative conditioning in contemporary digital ecosystems, where micro-targeted visual pairings subtly shape consumer desires beneath the threshold of conscious defense.

11.2 Prejudice, Stigmatization, and Social Cognitive Biases

Beyond commercial consumption, the evaluative conditioning paradigm has provided transformative insights into the social psychological origins of implicit prejudice, social stereotyping, and racial bias. Social cognitive researchers, notably Anthony Greenwald, Mahzarin Banaji, and Brian Nosek, demonstrated that human social preferences are profoundly influenced by implicit, automatic associations that frequently contradict an individual’s consciously stated, egalitarian beliefs.

Evaluative conditioning represents a primary cultural vector through which these implicit biases are established. Throughout an individual’s development, racial, ethnic, religious, and gender outgroups are frequently presented within media, cinematic, and cultural discourses alongside negatively valenced stimuli—such as visual depictions of crime, poverty, disease, aggression, or cultural mockery. Operating in exact accordance with the Martin and Levey picture-picture model, these repeated contiguous pairings drive a subtle, cumulative transfer of negative hedonic valence to the visual representation of outgroup faces. The outgroup facial cues become conditioned stimuli ($CS^-$) that trigger micro-affective reactions of discomfort, threat, or aversion.

To measure these conditioned social valences, researchers developed the Implicit Association Test (IAT), an empirical tool that quantifies the strength of associative links between social categories and positive or negative evaluative concepts. Recognizing that evaluative conditioning drives the formation of implicit bias, social psychologists have utilized Martin and Levey’s paradigm to design targeted counterconditioning interventions. By exposing individuals to extensive pairing arrays where marginalized social groups are systematically paired with intensely positive, admirable, and heroic unconditioned stimuli, researchers have successfully attenuated ingrained implicit social biases, demonstrating the clinical and societal utility of the evaluative architecture.

11.3 Etiology and Treatment of Phobias and Affective Disorders

Because Irene Martin and A.B. Levey were based at the Institute of Psychiatry and the Maudsley Hospital, their original theoretical focus was deeply intertwined with the etiology and treatment of clinical neurosis, irrational anxiety, and depressive disorders. Historically, behavioral clinical psychology explained phobias through John B. Watson’s classical fear conditioning model: an individual experiences a traumatic physical event (such as being bitten by a dog), wherein the dog becomes a CS signaling physical pain (US), producing a conditioned fear reflex.

However, psychiatric epidemiologists noted that the vast majority of individuals suffering from severe phobias, obsessive-compulsive cleaning rituals, or specific disgust triggers report zero historical trauma involving physical pain. An individual may possess an incapacitating phobia of spiders, snakes, or dirty public doorknobs without ever having suffered an actual physical injury from them. Martin and Levey argued that these neurotic conditions are frequently established via evaluative conditioning rather than classical fear conditioning. The neutral object was never paired with physical tissue damage; rather, it was contiguously paired with visceral, intensely negative visual and emotional stimuli—such as witnessing parental revulsion, viewing depictions of disease, or experiencing severe social disgust.

This evaluative etiology carries monumental clinical implications for exposure therapy. Traditional exposure therapy is founded on Pavlovian extinction: therapists repeatedly expose the patient to the phobic CS in a safe environment, expecting the conditioned fear response to extinguish as the patient learns the predictive contingency that “the spider does not bite.” However, clinicians have long struggled with the notorious phenomenon of phobic relapse and the stubborn persistence of clinical disgust. Martin and Levey’s paradigm explains this clinical failure: because evaluative conditioning is intrinsically resistant to extinction, simply showing a patient that a spider is harmless does not make the spider pleasant. The patient may cognitively accept that the spider is safe, yet the conditioned aesthetic revulsion and disgust remain completely unperturbed. Modern clinical behavior therapy has consequently adopted counterconditioning protocols—actively pairing phobic stimuli with deeply soothing, humorous, or pleasant stimuli—to rewrite the indelible evaluative substrate identified by Martin and Levey.

11.4 Health Behaviors, Addictive Appetites, and Aversive Conditioning

In the public health arena, evaluative conditioning principles have been mobilized to modify maladaptive health behaviors, combat addictive appetites, and combat global epidemics of obesity and tobacco addiction. A prime manifestation of this application is found in global tobacco regulation policies: the implementation of mandatory graphic cigarette warning labels.

Governments across dozens of nations mandate that cigarette packaging feature large, graphic photographic representations of gangrenous lungs, cancerous oral tumors, and post-mortem surgical autopsies. This policy is a direct real-world application of Martin and Levey’s visual picture-picture pairing paradigm. The cigarette brand logo and physical packaging serve as the neutral or positively valenced CS; the graphic medical photograph serves as a visceral, intensely repulsive unconditioned stimulus ($US^-$). Through repeated contiguous exposure every time a smoker reaches for a cigarette, evaluative conditioning degrades the positive hedonic valence of the brand and replaces it with conditioned disgust and visceral aversion, driving long-term smoking cessation.

A similar strategy is deployed in nutritional psychology to combat the consumption of high-calorie, processed junk foods. Researchers have designed smartphone applications and digital interventions that contiguously pair images of sugary sodas and hyper-processed foods with images of visceral decay, obesity-related pathology, or environmental pollution. Conversely, positive evaluative conditioning is utilized to encourage healthy eating in young children by systematically pairing visual depictions of raw vegetables with beloved cartoon characters or colorful, energetic imagery. By bypassing intellectual nutritional education and targeting the primal, automatic evaluative mechanisms mapped by Martin and Levey, public health researchers deploy associative learning to modify human behavior at scale.

12. The Enduring Legacy of Martin and Levey in Contemporary Affective Science

12.1 Neurobiological Correlates of Evaluative Learning

The dawn of twenty-first-century cognitive neuroscience provided the technological tools necessary to evaluate the physiological and neural architecture underlying the theoretical predictions made by Martin and Levey in 1975. Modern functional Magnetic Resonance Imaging (fMRI), positron emission tomography (PET), and high-density electroencephalography (EEG) have illuminated the specific neural circuits that govern evaluative acquisition and expression.

Neuroimaging paradigms directly adapting Martin and Levey’s picture-picture methodology have revealed that evaluative conditioning engages a distinct neurobiological network centered on the amygdala, the orbitofrontal cortex (OFC), and the anterior insular cortex. The orbitofrontal cortex is recognized as the primary cortical computational hub for encoding subjective hedonic value and tracking real-time changes in stimulus preference. When a neutral visual CS is paired with a liked or disliked US, neuroimaging tracks the progressive tuning of neurons within the OFC: as conditioning unfolds, OFC firing patterns to the CS alone begin to mimic the precise neural patterns originally evoked by the US.

Crucially, neuroscientific studies have confirmed the neural dissociation between contingency awareness and evaluative transfer predicted by Martin and Levey. Patients with localized damage to the hippocampus (the neurobiological substrate of conscious, declarative memory) are completely incapable of reporting which visual stimuli were paired together; yet, when presented with the CS, they demonstrate robust, normal evaluative shifts accompanied by differential activation of the amygdala and ventral striatum. Conversely, patients with focal damage to the amygdala or ventromedial prefrontal cortex can verbally state the pairings with perfect accuracy, yet display zero evaluative valence transfer. This double dissociation provides definitive neurobiological validation for Martin and Levey’s original assertion that evaluative conditioning is an autonomous learning system rooted in distinct affective substrates.

12.2 Epistemological Shift: Elevating Affect within Learning Theory

The lasting legacy of Irene Martin and A.B. Levey extends far beyond the empirical parameters of a single visual paradigm. Historically, their work catalyzed an epistemological revolution that fundamentally reshaped the landscape of experimental psychology. By having the intellectual courage to measure human subjective preference within an era dominated by rigid behaviorist reflexology and mechanical cognitivism, they demarginalized affect.

Before 1975, mainstream associative learning theory treated emotional feelings as secondary, untrustworthy epiphenomena—mere ripples on the surface of overt physiological reflexes. Martin and Levey demonstrated that subjective evaluative feelings could be measured with psychophysical rigor, obeyed systematic associative laws, and possessed their own unique structural properties. Their work broke the monopoly of the motor reflex, proving that learning involves not only learning what to do or what will happen next, but also learning how to feel.

This epistemological shift paved the way for the flourishing of modern affective neuroscience, appraisal theories of emotion, and behavioral economics. It forced cognitive science to abandon the sterile metaphor of the human mind as a purely logical computer, replacing it with an integrated view of human cognition where affect and association are inextricably intertwined. The methodological boldness of their Maudsley laboratory experiments proved that human aesthetic and emotional life—long thought to be the exclusive domain of the humanities or psychoanalytic speculation—could be unlocked through the tools of experimental psychological science.

12.3 Open Questions and Future Trajectories in Evaluative Conditioning

As evaluative conditioning approaches its sixth decade as an active empirical discipline, the theoretical questions first opened by Martin and Levey continue to evolve and inspire new generations of researchers. The contemporary landscape is characterized by intense computational modeling, as researchers utilize Bayesian updating models and connectionist neural networks to simulate the exact algorithmic rules that govern hedonic transfer in human memory networks.

Major frontiers of inquiry continue to push the boundaries of the paradigm:

• The Propositional-Associative Synthesis: Advanced computational cognitive science is moving beyond the binary “awareness versus non-awareness” debate, developing integrated models that explore how low-level associative binding mechanisms interface dynamically with high-level propositional reasoning.
• Developmental and Lifespan Trajectories: Developmental psychologists are tracking individual differences in evaluative conditionability across the human lifespan, investigating how the plasticity of affective learning changes from early infantile attachment through neurodegenerative decline in late adulthood.
• Digital and Algorithmic Environments: In an era dominated by artificial intelligence, predictive algorithms, and immersive virtual reality, understanding how human evaluative conditioning occurs within high-frequency digital architectures has become an urgent societal priority.

From its modest origins in an English psychiatric laboratory using physical art postcards, evaluative conditioning has grown into a cornerstone of psychological science. The landmark experiment of Irene Martin and A.B. Levey stands as an enduring monument to scientific innovation—a profound empirical demonstration of how the subtle co-occurrences of our physical world quietly, continuously, and permanently write the architecture of the human heart.

Conclusion

The story of Irene Martin and A.B. Levey’s evaluative conditioning experiment is a masterclass in scientific paradigm shifting. Working against the mechanistic grain of mid-1970s learning theory, they looked beyond the predictable twitches of human eyelids and the autonomic surges of skin conductance to capture something infinitely more elusive: the birth of a human preference. Through their elegant, deceptively simple “picture-picture” methodology, they demonstrated that our likes and dislikes are not purely innate, nor are they the exclusive product of conscious, rational deliberation. Instead, they are continuously sculpted by the associative currents of our environment, woven together by the spatiotemporal contiguity of the objects, images, and experiences that populate our lives.

Over fifty years of empirical scrutiny, theoretical debate, and cross-disciplinary application have affirmed the brilliance of their initial insights. Evaluative conditioning has redefined the boundaries of associative learning, proved that affect possesses its own autonomous cognitive architecture, and provided modern society with the empirical tools to understand consumer culture, dismantle social prejudices, and treat chronic psychiatric distress. Irene Martin and A.B. Levey did not simply introduce a new experimental paradigm; they illuminated the silent, associative machinery that transforms the raw canvas of environmental perception into the rich, qualitative tapestry of human emotional experience.

References

• Baeyens, D., Eelen, F., & Crombez, G. (1988). Pavlovian conditioning of visual-evaluative values in humans. Cognition and Emotion, 2(4), 309–322. https://doi.org/10.1080/02699938808412702
• Baeyens, D., Eelen, F., Crombez, G., & Van den Bergh, O. (1992). Human evaluative conditioning: Acquisition trials, presentation schedule, and contingency awareness. Behaviour Research and Therapy, 30(2), 133–142. https://doi.org/10.1016/0005-7967(92)90135-N
• Clore, G. L., & Schwarz, N. (1999). The psychobiology of emotion: The role of affect in information processing. In M. D. Power & T. Dalgleish (Eds.), Handbook of Cognition and Emotion (pp. 43–64). John Wiley & Sons. https://doi.org/10.1002/0470013494.ch3
• De Houwer, J., Thomas, S., & Baeyens, D. (2001). Association learning of likes and dislikes: A review of 25 years of research on human evaluative conditioning. Psychological Bulletin, 127(6), 853–869. https://doi.org/10.1037/0033-2909.127.6.853
• De Houwer, J. (2007). A propositional approach to evaluative learning. In F. De Houwer & D. Hermans (Eds.), Cognition and Emotion: A Broad View (pp. 244–266). Psychology Press. https://doi.org/10.4324/9780203837948
• Eysenck, H. J. (1967). The Biological Basis of Personality. Charles C. Thomas.
• Field, A. P. (2000). Evaluative conditioning is sensitive to contingency awareness: Do will-o’-the-wisps exist? Cognition and Emotion, 14(4), 583–600. https://doi.org/10.1080/026999300402802
• Hofmann, W., De Houwer, J., Perugini, M., Baeyens, D., & Crombez, G. (2010). Evaluative conditioning in humans: A meta-analysis. Psychological Bulletin, 136(3), 390–421. https://doi.org/10.1037/a0018916
• Hollands, J. G., Rozin, P., & Fallon, A. (1989). Evaluative conditioning with disgust-eliciting stimuli. Unpublished Manuscript, University of Pennsylvania.
• Martin, I., & Levey, A. B. (1969). The genesis of the classical conditioned response. International Review of Neurobiology, 11, 267–320. https://doi.org/10.1016/S0074-7742(08)60388-7
• Martin, I., & Levey, A. B. (1975). Evaluative conditioning. Behaviour Research and Therapy, 13(4), 211–221. https://doi.org/10.1016/0005-7967(75)90025-8
• Martin, I., & Levey, A. B. (1978). Evaluative conditioning: A survey of some initial findings. Behaviour Research and Therapy, 16(4), 283–289. https://doi.org/10.1016/0005-7967(78)90025-4
• Martin, I., & Levey, A. B. (1987). Learning what will happen next: Conditioning and thinking. In H. J. Eysenck & I. Martin (Eds.), Theoretical Foundations of Behaviour Therapy (pp. 53–81). Plenum Press. https://doi.org/10.1007/978-1-4757-9438-0_4
• Osgood, C. E., Suci, G. J., & Tannenbaum, P. H. (1957). The Measurement of Meaning. University of Illinois Press.
• Rescorla, R. A. (1968). Probability of shock in the presence and absence of CS in fear conditioning. Journal of Comparative and Physiological Psychology, 66(1), 1–5. https://doi.org/10.1037/h0025984
• Rozin, P., & Zellner, D. (1985). The role of Pavlovian conditioning in the acquisition of food likes and dislikes. Annals of the New York Academy of Sciences, 443(1), 189–202. https://doi.org/10.1111/j.1749-6632.1985.tb27073.x
• Shanks, D. R., & Dickinson, A. (1990). Contingency awareness in evaluative conditioning: A commentary on Baeyens et al. Cognition and Emotion, 4(1), 19–30. https://doi.org/10.1080/02699939008406763
• Staats, A. W., & Staats, C. K. (1958). Attitudes established by classical conditioning. The Journal of Abnormal and Social Psychology, 57(1), 37–40. https://doi.org/10.1037/h0042782
• Walther, E., Nagengast, B., & Trasselli, C. (2005). Evaluative conditioning in social psychology: Facts and speculations. Cognition and Emotion, 19(2), 175–196. https://doi.org/10.1080/02699930441000274
• Zajonc, R. B. (1980). Feeling and thinking: Preferences need no inferences. American Psychologist, 35(2), 151–175. https://doi.org/10.1037/0003-066X.35.2.151