4.1 Classical Conditioning: Paradigms, Extinction, Spontaneous Recovery, and Biological Constraints

Classical Conditioning: Paradigms, Extinction, Spontaneous Recovery, and Biological Constraints

Classical conditioning—first systematically characterized by Russian physiologist Ivan Pavlov—represents a foundational mechanism by which organisms learn predictive relationships between environmental events. For the GRE Subject Test in Psychology, students must master not only the classical nomenclature and temporal arrangements but also modern associative learning theories, formal mathematical models, and the evolutionary boundaries that constrain conditioning.

---

1. The Pavlovian Conditioning Paradigm and Terminology

While investigating the neural regulation of canine digestion using chronic salivary fistulas, Pavlov observed that dogs began salivating prior to the physical delivery of meat powder—responding to the sight of the food bowl, the sound of approaching laboratory footsteps, or the appearance of the research assistant. Pavlov termed these anticipatory secretions 'psychic secretions,' recognizing that they reflected learned reflexes.

The Four Core Components

Classical conditioning involves four definitive elements:

1. Unconditioned Stimulus (US): An environmental event that unconditionally, reflexively, and reliably elicits an innate biological reaction without prior training (e.g., meat powder on the tongue, an electric shock, a corneal puff of air).
2. Unconditioned Response (UR): The unlearned, innate physiological or behavioral reflex automatically triggered by the US (e.g., salivation, leg flexion, eyeblink).
3. Conditioned Stimulus (CS): An initially neutral stimulus that does not evoke the target reflex. Through systematic, contingent temporal pairing with the US, the CS comes to evoke an acquired response (e.g., a 1000-Hz auditory tone, a metronome click, a visual light flash).
4. Conditioned Response (CR): The learned behavioral or physiological response elicited by the CS alone following conditioning (e.g., salivation upon hearing the tone).

Before Conditioning:
  CS (Tone) ────────────► Neutral Response (Orientation, no salivation)
  US (Meat Powder) ─────► UR (Salivation)

During Conditioning (Acquisition):
  CS (Tone) + US (Meat Powder) ─────► UR (Salivation)

After Conditioning:
  CS (Tone) ────────────► CR (Salivation)

Stimulus-Substitution vs. Preparatory-Response Theories

Pavlov advanced the Stimulus-Substitution Theory, proposing that conditioning establishes direct functional neural connections between the cortical centers processing the CS and those processing the US. In this view, the CS becomes a literal neural surrogate for the US, predicting that the CR must always be physiologically identical to the UR.

Subsequent research disproved literal stimulus substitution. Jerzy Konorski and later Gregory Kimble demonstrated the Preparatory-Response Theory, showing that the CR prepares the biological system for the impending US. In many paradigms, the CR is qualitatively different from, or even opposite to, the UR:

• Conditioned Compensatory Responses (Siegel): In morphine administration, the direct pharmacological effect (UR) includes hypothermia, analgesia, and respiratory depression. However, environmental cues consistently associated with drug administration (CS) elicit homeostatic, compensatory responses (CR): hyperthermia and hyperalgesia. Shepard Siegel demonstrated that conditioned compensatory responses explain drug tolerance and lethal overdose: when experienced heroin users self-administer their typical dose in a novel environment lacking familiar contextual CSs, the absence of compensatory anticipatory physiological adjustments leaves the organism unprotected against fatal respiratory depression.

---

2. Temporal Arrangements of CS and US

The temporal relationship between the onset of the CS and the delivery of the US dramatically determines the rate of acquisition and the asymptotic strength of the CR. The interval between CS onset and US onset is designated the interstimulus interval (ISI).

1. Short-Delay:     CS  [==============]
                    US            [====]

2. Long-Delay:      CS  [===========================]
                    US                        [====]

3. Trace:           CS  [=======]
                                  (Trace Gap)
                    US                       [====]

4. Simultaneous:    CS  [==============]
                    US  [==============]

5. Backward:        US  [==============]
                    CS            [====]

Detailed Analysis of Paradigms

---

3. Extinction, Recovery Phenomena, and the Non-Erasure Principle

When a conditioned stimulus is repeatedly presented in the complete absence of the unconditioned stimulus ($CS\text{-alone}$ trials), the conditioned response progressively declines in frequency and amplitude until reaching baseline levels. This process is termed extinction.

Extinction is New Inhibitory Learning, Not Unlearning

A fundamental tenet of contemporary learning theory is that extinction does not erase the original excitatory CS-US associative trace. Instead, extinction represents active, new inhibitory learning: the organism forms a context-dependent inhibitory association ($CS\text{-noUS}$) that superimposes upon and suppresses the underlying excitatory memory ($CS\text{-US}$). Five classic empirical phenomena unequivocally demonstrate that the original associative memory survives extinction:

1. Spontaneous Recovery: Following successful extinction, if the subject is removed from the experimental chamber and returned after a retention interval (e.g., 24 hours), presentation of the CS alone reliably elicits a temporary return of the CR without any retraining. The magnitude of spontaneous recovery is proportional to the duration of the rest interval.
2. The Renewal Effect (Mark Bouton): Bouton demonstrated that extinction is highly context-specific, whereas acquisition generalizes across contexts. When an extinguished CS is encountered outside the specific extinction context, conditioned responding abruptly re-emerges:
   • ABA Renewal: Acquisition occurs in Context A; extinction occurs in Context B; testing in Context A produces massive renewal of the CR.
   • ABC Renewal: Acquisition in Context A; extinction in Context B; testing in a novel Context C yields significant renewal.
   • AAB Renewal: Acquisition and extinction both occur in Context A; testing in Context B yields renewal.
3. Reinstatement: Following extinction, non-contingent exposure to the US alone in the testing environment—completely unpaired with the CS—subsequently restores conditioned responding to the CS when it is reintroduced.
4. Disinhibition: During an extinction session where the CR has diminished, presenting a novel, extraneous stimulus (e.g., a loud clicking sound or novel flash) immediately alongside the extinguished CS causes an instantaneous, transient recovery of the CR on that trial. The novel stimulus disrupts the fragile, active cortical inhibition established during extinction.
5. Rapid Reacquisition: When an extinguished CS is once again paired with the US, the rate of acquisition is vastly accelerated compared to initial naive conditioning, demonstrating savings of the underlying associative trace.

---

4. Generalization, Discrimination, and Associative Networks

Stimulus Generalization and Gradients

Once a CR is established to a specific CS, stimuli physically similar to the trained CS also elicit the CR—a phenomenon known as stimulus generalization. In classic experiments by Guttman and Kalish (1956), pigeons reinforced for pecking a visual key illuminated with a 580-nm wavelength light exhibited a bell-shaped generalization gradient when tested across varying wavelengths under extinction. The peak of responding occurred at 580 nm, with systematic, symmetrical declines in response rate as the physical test stimulus deviated further along the nanometer spectrum.

        Response
          Rate
           ▲
           │              ▲ [CS: 580 nm]
           │             / \
           │            /   \
           │           /     \
           │          /       \
           │        ─┘         └─
           └──────────────────────────►
            520   540   560   580   600   620   640  (Wavelength in nm)

Stimulus Discrimination and Peak Shift

Stimulus discrimination occurs when an organism learns to respond selectively to one stimulus ($CS+$) paired with the US, while withholding responses to a distinct stimulus ($CS-$) presented without the US. Hanson (1959) discovered the peak shift phenomenon: when animals receive intradimensional discrimination training (e.g., $CS+$ at 550 nm, $CS-$ at 560 nm), the peak of the generalization gradient shifts away from the $CS-$ in the direction opposite to the non-reinforced stimulus (e.g., peaking at 540 nm instead of the trained 550 nm), reflecting the interaction between an excitatory gradient centered on $CS+$ and an inhibitory gradient centered on $CS-$.

Higher-Order Conditioning vs. Sensory Preconditioning

Associative learning can occur without direct contact between a stimulus and a biological US:

Higher-Order Conditioning:                     Sensory Preconditioning:
Phase 1: CS1 ──► US                            Phase 1: CS2 ──► CS1 (No US)
Phase 2: CS2 ──► CS1                           Phase 2: CS1 ──► US
Test:    CS2 ──► CR                            Test:    CS2 ──► CR

• Higher-Order (Second-Order) Conditioning: A neutral stimulus ($CS_1$) is first paired with a biological US until it reliably evokes a CR. In Phase 2, a novel neutral stimulus ($CS_2$) is repeatedly paired with $CS_1$ in the total absence of the US. In testing, $CS_2$ successfully elicits the CR. Third-order conditioning is possible but notoriously difficult to establish because Phase 2 trials simultaneously act as extinction trials for $CS_1$.
• Sensory Preconditioning (Brogden, 1939): In Phase 1, two neutral stimuli ($CS_2$ and $CS_1$) are repeatedly paired together without any unconditioned stimulus (e.g., a tone is followed by a light). In Phase 2, $CS_1$ alone is paired with a biological US (e.g., shock) until a robust CR develops. In the test phase, $CS_2$ is presented alone. Despite never having been paired with the US or an established conditioned stimulus, $CS_2$ elicits the CR. Sensory preconditioning proves that organisms acquire silent, stimulus-stimulus (S-S) associations during passive environmental co-occurrence in the absence of reinforcement.

---

5. The Rescorla-Wagner Mathematical Model of Conditioning

In 1972, Robert Rescorla and Allan Wagner formalized classical conditioning as a mathematical function of prediction error: learning occurs not simply because a CS and US co-occur in time, but because the delivery of the US is surprising relative to the organism's current expectations.

The Mathematical Formulation

The change in associative strength of a given stimulus $A$ on trial $n$ is defined by:

$\Delta V_A = \alpha_A \beta_X (\lambda - \Sigma V)$

Where:

• $\Delta V_A$: The change in the associative strength of conditioned stimulus $A$ on that specific trial.
• $\alpha_A$: The salience or perceptual prominence of conditioned stimulus $A$ ($0 \le \alpha \le 1$).
• $\beta_X$: The learning rate parameter determined by the nature and intensity of the US ($0 \le \beta \le 1$).
• $\lambda$ (lambda): The maximum associative strength that the specific US can support (the asymptote of conditioning, representing actual US magnitude; $\lambda = 0$ on extinction trials).
• $\Sigma V$: The sum of the associative strengths of all conditioned stimuli present on that trial (the organism's total expectation of the US).
• $(\lambda - \Sigma V)$: The prediction error. When the organism's expectation matches reality ($\Sigma V = \lambda$), prediction error becomes zero, and learning terminates.

Explanatory Triumphs of the Rescorla-Wagner Model

Theoretical Limitations of Rescorla-Wagner

Despite its mathematical elegance, Rescorla-Wagner cannot account for several verified conditioning phenomena:

• Latent Inhibition (CS Pre-Exposure Effect): Presenting a CS repeatedly by itself prior to conditioning retards subsequent excitatory conditioning. Because $\Sigma V = 0$ and $\lambda = 0$ during pre-exposure, Rescorla-Wagner predicts $\Delta V = 0$, falsely expecting no effect of pre-exposure (addressed by Nicholas Mackintosh's attentional model and the Pearce-Hall model of stimulus associability).
• Spontaneous Recovery and Renewal: Rescorla-Wagner assumes extinction reduces associative strength toward zero, making it mathematically impossible to predict the spontaneous return of conditioned responding without new training.

---

6. Biological Preparedness and Constraints on Learning

For decades, radical behaviorists operated under the equipotentiality premise (or general-process learning theory)—the assumption that the laws of conditioning apply uniformly to any stimulus and any response, regardless of species or ecological niche. In Pavlov's words, 'Any natural phenomenon chosen at will may be made a conditioned stimulus.' This dogma was definitively shattered in the 1960s.

Garcia & Koelling's 'Bright-Noisy Water' Experiment (1966)

John Garcia and Robert Koelling exposed thirsty rats to drinking spouts that simultaneously delivered three sensory cues: a distinctive gustatory taste (saccharin sweet water) and compound audiovisual cues (flashing lights and auditory clicks activated by licking). Rats were divided into two experimental groups subjected to different unconditioned stimuli:

• Group 1 (Visceral Illness US): Injected with lithium chloride ($LiCl$) or exposed to X-ray irradiation, producing delayed internal gastric malaise.
• Group 2 (Cutaneous Pain US): Received immediate painful peripheral electric foot-shocks while drinking.

CS Compound: [Taste (Saccharin) + Audiovisual (Bright-Noisy)]
                        │
       ┌────────────────┴────────────────┐
       ▼                                 ▼
[US: Lithium Chloride Illness]    [US: Electric Foot-Shock]
       │                                 │
       ▼                                 ▼
Avoided: TASTE                    Avoided: BRIGHT-NOISY
Ignored: Bright-Noisy             Ignored: Taste

When tested on the separate components, rats made sick by $LiCl$ exhibited profound avoidance of the taste, but drank normally in the presence of the bright-noisy cues. Conversely, rats that received electric shock showed intense fear of the bright-noisy cues, but drank saccharin-flavored water without hesitation.

This double dissociation proved selective association: organisms are biologically predisposed to connect internal visceral distress with gustatory inputs (internal defense system) and peripheral physical trauma with external audiovisual inputs (external defense system).

Conditioned Taste Aversion (CTA / The Garcia Effect)

Conditioned Taste Aversion exhibits unique physiological parameters that violate the traditional operational canons of classical conditioning:

1. One-Trial Learning: A robust, long-lasting conditioned taste aversion can develop after a single pairing of a novel taste and visceral illness.
2. Extreme CS-US Delays: Unlike motor conditioning paradigms that degrade within seconds, CTA develops even when illness follows consumption by several hours (up to roughly 24 hours in some rat studies).
3. Biological Preparedness: Martin Seligman categorized learning along a continuum of preparedness based on evolutionary adaptiveness:
   • Prepared Associations: Innately wired neural circuits allow rapid, highly resistant conditioning with minimal input (e.g., taste-illness, phobias to snakes and spiders).
   • Unprepared Associations: Arbitrary environmental pairings that require extensive, multi-trial exposure (e.g., tone-shock, geometric patterns paired with food).
   • Contraprepared Associations: Pairings that run directly counter to an animal's evolutionary adaptations and are nearly impossible to acquire (e.g., nausea paired with audiovisual cues in rodents).