Basic principles

Pavlovian conditioning was originally discovered by Pavlov (e.g., Pavlov, 1927). In his research on the physiology of dog’s digestion, he observed that by repeatedly pairing an initially neutral stimulus (e.g., a metronome sound) with the delivery of food, these animals started to salivate in response to the stimulus, rather than merely salivating in the presence of the food. Pavlov interpreted this finding as reflecting the fact that the stimulus progressively acquired the ability to elicit an anticipatory salivary response through its systematic association with the food.

Formally, Pavlovian conditioning corresponds to the learning process and procedure by which an environmental stimulus (the conditioned stimulus, CS) is associated with a motivationally salient aversive or appetitive event (the unconditioned stimulus, US) by virtue of a single or repeated contingent pairing (Fanselow & Wassum, 2016; Pavlov, 1927; Rescorla, 1988b; see Figure 2.1). Usually (but not necessarily), the CS is an initially neutral stimulus (e.g., a tone or a light) that does not evoke a specific response (except orienting in some situations) prior to conditioning; by contrast, the US is biologically potent (e.g., electric stimulation or pleasant food) and automatically triggers an emotional response (the unconditioned response, UR) without prior learning, such as physiological reactions (e.g., increased skin conductance, heart rate, blood pressure, or salivation). After its pairing with the US, the organism learns that the CS predicts the US, and the presentation of the CS alone produces a conditioned response (CR), which often encompasses a constellation of emotional reactions (e.g., Phelps, 2006). These reactions can generally be observed at the behavioral (e.g., freezing, avoidance or approach behaviors), physiological (e.g., stress hormone release, increased startle responses, skin conductance, heart rate, or salivation), neural (e.g., increase in BOLD signal in the amygdala), and subjective (e.g., feelings of fear or pleasure) levels (e.g., LeDoux, 2012; Phelps, 2006). Importantly, the responses evoked by the CS are not identical to those elicited by the US. Albeit generally similar, the CR and the UR may differ, or even be opposite in certain cases. For instance, studies on conditioned analgesia have shown that whereas an electric stimulation (i.e., the US) typically elicited freezing behaviors in rats (i.e., the UR), the presentation of the tone that was established as a predictive cue for the electric stimulation (i.e., the CS) triggered endorphin release (i.e., the CR), thereby preparing the organism to cope with the upcoming electric stimulation by diminishing their pain sensitivity,

which ultimately resulted in a reduction in their amount of freezing (e.g., Fanselow & Bolles, 1979). This result notably indicates that the key mechanism in Pavlovian conditioning pertains to the association between the CS and the US rather than the CS substituting for the US in eliciting the UR in a reflexive manner as initially proposed by Pavlov (1927).

CS-US contingency

Correspondingly, the relation between the CS and the US is a central aspect of Pavlovian conditioning. Whereas the mere co-occurrence or contiguity between the CS and the US was initially considered as the determining condition for Pavlovian learning (e.g., Pavlov, 1927), it has been demonstrated that such contiguity is neither sufficient nor necessary. Instead, Pavlovian conditioning only occurs if the CS has a predictive relationship with the US (i.e., the presentation of the US is contingent upon the occurrence of the CS; Rescorla, 1967, 1988b).

Figure 2.1. Basic principles of Pavlovian conditioning. Through a single or repeated contingent pairing with the unconditioned stimulus (US) that automatically triggers an unconditioned response (UR) without prior learning, a typically (but not necessarily) initially neutral stimulus (NS) becomes conditioned (conditioned stimulus, CS), thereby acquiring an emotional and predictive value eliciting a preparatory response (conditioned response, CR).

Unconditioned


For instance, the importance of the CS-US contingency is illustrated by Rescorla’s (1968) experiment, in which he trained three groups of rats to associate a tone CS with an electric stimulation as the US. In the first group, the probability of the US was greater when the CS was presented than when it was absent (i.e., positive contingency). In the second group, the probability of the US was the same whether or not the CS was presented (i.e., zero contingency), the number of CS-US pairings being equivalent as in the first group and additional US presentations being delivered during the intertrial interval. In the third group, the probability of the US was lower when the CS was presented than when it was absent (i.e., negative contingency). Results showed that excitatory conditioning to the CS occurred in the first group, while inhibitory conditioning happened in the third group, the CS becoming a safety signal. By contrast, no conditioning was observed in the second group, even though an identical number of CS-US pairings was received as in the first group, which indicates that Pavlovian conditioning only occurred when there was a contingency relationship between the CS and the US.

Another example of the CS-US contingency relevance is the blocking effect (Kamin, 1968, 1969). Blocking refers to the finding that the conditioning of the association between a conditioned stimulus, CSA, and the US is impaired if CSA is presented in compound with another conditioned stimulus, CSB, that has been previously associated with the US during conditioning trials, CSB “blocking” conditioning to CSA (Bouton, 2007; Fanselow & Wassum, 2016; Kamin, 1969). This effect suggests that simply pairing a CS with a US is not sufficient for producing Pavlovian conditioning, and that conditioning is achieved only when a CS has informational value (i.e., it provides new information about the US beyond what is already predicted by other CSs; Bouton, 2007). It is worth noting that explicit awareness of the contingency between the CS and the US is, however, likely not necessary for Pavlovian conditioning to occur in animals (see, e.g., Papini & Bitterman, 1990) and in humans (e.g., Bechara et al., 1995), although the specific role of contingency awareness in humans remains debated (e.g., Lovibond & Shanks, 2002; Öhman & Mineka, 2001).

Extinction

In addition to acquisition (i.e., learning resulting from the association between the CS and the US), an essential phenomenon observed in Pavlovian conditioning is extinction. It occurs when the US is no longer delivered after the CS and/or when the CS is no longer predictive of the US, which results in a gradual weakening or decrease in the probability of the CR occurrence over time. Extinction refers both to the procedure of presenting the CS in

absence of the US and to the phenomenon resulting from that procedure. It constitutes a crucial process in behavior change (e.g., Bouton, 2007; Dunsmoor, Niv, et al., 2015). Extinction indeed allows the organism to adapt to a changing environment by stopping producing responses and behaviors that are no longer reinforced. Extinction processes also have a high clinical significance for the treatment of a variety of psychiatric conditions (e.g., Dunsmoor, Niv, et al., 2015; Milad & Quirk, 2012), as well as serve as the basis for exposure therapy, which is one of the most effective treatment for anxiety disorders, phobias, stress-related disorders, and addictions (see, e.g., Craske, Treanor, Conway, Zbozinek, & Vervliet, 2014;

Dunsmoor, Niv, et al., 2015).

Importantly, extinction has been theorized to involve unlearning or erasure of the original CS-US association (e.g., Rescorla & Wagner, 1972), or, alternatively, to induce new inhibitory learning between the CS and the US, the CS thereby acquiring inhibitory properties that reduce or suppress the CR (e.g., Bouton, 2002; Bouton, Westbrook, Corcoran, & Maren, 2006; Konorski, 1967; Pearce & Hall, 1980). In line with the latter view, research on extinction has unveiled the existence of at least four phenomena suggesting that original learning (i.e., acquisition) is not merely erased: (1) spontaneous recovery, which consists of the recovery of the CR when the CS is tested after time has passed following the end of extinction training, (2) reinstatement, which refers to the CR recovery occurring after the organism is exposed to the US alone after extinction, (3) renewal, which corresponds to the CR recovery that can happen when the context is changed after extinction, and (4) rapid reacquisition, which pertains to the rapid CR recovery when the CS is paired with the US again after extinction (Bouton, 2002).

Nevertheless, recent theoretical advances in the study of computational mechanisms underlying extinction processes have suggested that extinction may induce both the formation of new inhibitory learning that interferes with the original excitatory CS-US association and attenuation or updating of this association with new information (Dunsmoor, Niv, et al., 2015;

Gershman, Blei, & Niv, 2010; Gershman & Hartley, 2015; Gershman, Monfils, Norman, &

Niv, 2017; Gershman & Niv, 2012).

Further, extinction learning is more fragile than the original CS-US association acquired through Pavlovian conditioning, as well as more sensitive to contextual information (Bouton, 2002; Bouton et al., 2006; Dunsmoor, Niv, et al., 2015). The asymmetry between extinction fragility and conditioning strength and persistence is however likely adaptive:

Although signals for danger or reward may not always, or even only rarely, contingently co-occur with an actual threatening or rewarding event in the environment, producing a rapid

defensive or approach response, respectively, remains critical in promoting survival when the occurrence of threat or reward is still possible (Dunsmoor, Niv, et al., 2015). Maintaining a memory trace of the original CS-US association may therefore help prepare the organism against the remote possibility of future threat or reward (Dunsmoor, Niv, et al, 2015).