The Nuances of Associative Learning: Beyond Direct Pairings
Associative learning, a cornerstone of behavioral science, describes the process by which individuals form connections between different stimuli or between a stimulus and a behavior. This fundamental form of learning underpins our ability to navigate the world, make predictions, and adapt our actions accordingly. While often considered to require the direct, physical presence of stimuli in the environment for them to be linked, this is not a necessary condition for learning to occur. Indeed, associative relationships can form between events that are never directly paired, highlighting the dynamic and integrative nature of memory.
Understanding the Foundations of Association
At its core, associative learning involves creating links between pieces of information. This can manifest in two primary categories: classical conditioning and operant conditioning.
Classical Conditioning: This form of learning occurs when an individual learns to associate a neutral stimulus with a significant event. The most famous example is Pavlov’s dogs, where the dogs learned to associate the sound of a bell with food, eventually responding to the bell alone by salivating. In the laboratory, this learning, often termed first-order conditioning (Pavlov, 1927), is modeled using Pavlovian conditioning. This process consists of pairings between a neutral sensory cue (or stimulus) such as a tone with an event of biological significance, such as an appetitive or aversive unconditioned stimulus (US). For instance, the painful experience of having been bitten by a dog can result in the development of fear of dogs, causing one to avoid places where dogs can be encountered. While this form of learning accurately captures the formation of many associative relationships, it misses many others.
Operant Conditioning: In operant conditioning, a behavior is associated with a consequence, which could either reinforce or discourage that behavior. For example, a person who is rewarded for working hard (with praise or a bonus) is more likely to repeat that behavior in the future. This is a "learning" or "conditioning" term that refers to learning that two different events occur or happen together. This is really a fundamental component of conditioning since a response to a stimulus won't really be learned if the organism doesn't get the point that the stimulus and response are supposed to occur together. This doesn't have to be a conscious learning, but the association must be made for the learning to occur. For example, will a rat learn to press a lever if it never makes the association between pressing the lever and getting the reward? Or why would a dog salivate to a bell if it never makes the connection between the bell and getting food?
Both types of associative learning shape how we respond to the world around us, affecting everything from our daily habits to our emotional reactions.
Read also: The Science of Associative Learning
Higher-Order Conditioning: Learning Through Indirect Association
While first-order conditioning demonstrates learning through direct pairings, many associative relationships are formed through more complex, indirect pathways. One need not directly experience event relationships in order to infer the likelihood of their occurrence in novel situations. To return to the example of the dog bite, one will likely not only avoid dogs (the stimulus directly associated with the aversive event) but also places where dogs frequent (e.g., parks, trails, your next-door neighbour’s yard) even though the bite had not occurred there. Here, the knowledge of where dogs can be encountered is integrated with the knowledge that dogs can cause painful bites. In other words, information acquired across different episodes or time points can be linked, thus offering an opportunity to infer unique event relationships and make novel predictions about the environment. Such integration is an example of the dynamic nature of memories, how memories become linked, and how flexible behaviour is orchestrated (Holland, 1990; Gewirtz and Davis, 2000; Blaisdell, 2009; Seitz et al., 2021).
This integration of distinct associative memories is elegantly captured in higher-order conditioning (Pavlov, 1927; Brogden, 1939). This learning consists of two conditioning episodes: one that leads to associative links between two neutral stimuli (e.g., S2→S1, where S2 could be an auditory cue such as a tone and S1 could be a visual cue such as a light), and another that links one of these stimuli (S1) with a biologically significant outcome (an appetitive or aversive US, i.e., S1→US). Subsequent presentations of S2 reveal its ability to invigorate conditioned responses (CRs) indicative of expectation of the US. This form of learning is termed higher-order because S2 is never directly paired with the US. Rather, it engages conditioned responding by virtue of its pairing with S1, which was directly associated with the US. That is, S2 acquires value through an intermediary.
Sensory Preconditioning vs. Second-Order Conditioning: A Matter of Order
There are two classic designs of higher-order conditioning: sensory preconditioning and second-order conditioning. While both types of designs consist of the same learning phases - sensory training and appetitive or aversive conditioning - the order of these phases is reversed.
- Sensory Preconditioning: In this design, S2→S1 pairings precede S1→US pairings. This means that two neutral stimuli are first paired together, and then one of those stimuli is paired with an unconditioned stimulus.
- Second-Order Conditioning: In contrast, second-order conditioning involves S1→US pairings preceding S2→S1 pairings. Here, a neutral stimulus is first paired with an unconditioned stimulus, and then that conditioned stimulus is paired with a new neutral stimulus.
Although the order of the learning phases may seem like a minor difference in experimental design, it is of tremendous importance because it governs what is learned during these distinct forms of higher-order conditioning. Accounts of higher-order learning were originally reported by Pavlov (1927), where cues directly paired with an appetitive or aversive outcome could support the acquisition of secondary conditioned reflexes when paired with novel cues in the absence of the associated outcome (i.e., second-order conditioning). In both humans and animals, Prokofiev and Zeliony (1926) reported that sensory pairings between two cues followed by aversive conditioning of one of those cues led to fear of the other (indirectly paired) sensory cue (i.e., sensory preconditioning). This was subsequently investigated more thoroughly by Brogden (1939), who coined the term “sensory preconditioning.”
These forms of higher-order learning have been replicated numerous times across species including drosophila, goldfish, pigeons, mice, rats, rabbits, monkeys, and humans (Reid, 1952; Rizley and Rescorla, 1972; Rashotte et al., 1977; Pfautz et al., 1978; Rescorla, 1979; Amiro and Bitterman, 1980; Cook and Mineka, 1987; Beauchamp and Gluck, 1988; Gibbs et al., 1991; Müller et al., 2000; Brembs and Heisenberg, 2001; Mead and Stephens, 2003; Tabone and de Belle, 2011; Lee and Livesey, 2012; Busquets-Garcia et al., 2017; Renaux et al., 2017; Craddock et al., 2018; Wong and Pittig, 2022). However, the precise design parameters employed can easily influence the strength and content of learning.
Read also: Comprehensive Guide to Subsidized Student Loans
Factors Influencing Higher-Order Conditioning
A variety of design factors have been reported in the literature to influence higher-order conditioning:
- Stimulus Arrangement: A simultaneous arrangement of stimuli results in a superior sensory preconditioning effect relative to a serial arrangement (Thompson, 1972; Rescorla, 1980b; Holland and Ross, 1983). This can be explained by considering the associations that form between the cues during sensory training. Simultaneous presentations facilitate associations between the sensory characteristics of S2 and S1, rather than a predictive relationship between them (i.e., S2 predicts S1 presentation), which is favored by a serial arrangement. Second-order conditioning can also be achieved using both simultaneous and serial S2 and S1 presentations. In studies where cues have been presented serially in sensory preconditioning or second-order conditioning, it is common for S2 to precede S1. However, instances of S1 preceding S2 (i.e., S1→S2) are also effective in supporting learning. In an aversive design, sensory preconditioning was successfully obtained using such a serial backward order (i.e., S1→S2; Ward-Robinson and Hall, 1998).
- Stimulus Similarity: Pairing of similar stimuli proceeds more rapidly relative to dissimilar stimuli in second-order conditioning (Rescorla and Furrow, 1977; Rescorla and Cunningham, 1979). Specifically, when similar stimuli are used in the roles of S2 and S1, higher-order conditioning is facilitated compared to using dissimilar stimuli. Rescorla and Furrow (1977) showed that second-order conditioning proceeded more rapidly when S1 and S2 belonged to the same, compared to different, class of stimuli (e.g., color: blue or green; orientation: horizontal or vertical lines). These effects were not due to stimulus generalization or pseudo-conditioning (Rescorla and Furrow, 1977). Cue similarity also facilitates second-order conditioning when the cues form a part-whole relationship. For example, in a pigeon autoshaping design, Rescorla (1980a) used achromatic shapes (triangle or square) as S2 and red shapes (triangle or square) as S1. Congruency in the shape, that is, when the achromatic shape was the same as the colored shape, resulted in better second-order conditioning.
- Number of Trials: Higher-order conditioning designs classically use forward serial pairings (Pavlov, 1927). Sensory preconditioning and second-order conditioning can be obtained in single S2 and S1 pairing in conditioned aversion preparation (Archer and Sjöden, 1982). Aversive higher-order learning proceeds in four trials for second-order conditioning and eight trials for sensory preconditioning (Parkes and Westbrook, 2010). Higher-order fear conditioning progresses fairly rapidly: four trials of serial S2→S1 pairings is sufficient to obtain second-order learning (Rizley and Rescorla, 1972; Parkes and Westbrook, 2010; Lay et al., 2018), and sensory preconditioning can be achieved in eight serial S2→S1 trials (Rizley and Rescorla, 1972; Parkes and Westbrook, 2011; Wong et al., 2019). Higher-order conditioning designs involving rewards require more extensive S2→S1 training. The large number of trials often required for second-order conditioning can have unintended effects. As the number of S2→S1 trials increase in second-order conditioning, responding to S2 decreases, which is in contrast with the increase in responding to S1 across S1→US pairings. When S2→S1 pairings are alternated with continued S1→US pairings, S2 can become a signal for the absence of the US (Herendeen and Anderson, 1968; Rescorla et al., 1973; Holland and Rescorla, 1975b; Yin et al., 1994). That is, conditioned inhibition to S2 accrues, competing with its ability to exhibit second-order conditioning (Gewirtz and Davis, 2000; Parkes and Westbrook, 2010). In a lick suppression study in rats, 20 simultaneous S2→S1 pairings favored conditioned inhibition over second-order conditioning, and a hundred such trials rendered S2 a conditioned inhibitor regardless of whether S2 and S1 were paired simultaneously or serially (Stout et al., 2004). The transition of S2 from a second-order excitor to a conditioned inhibitor was quicker when S2 and S1 were presented in compound (Stout et al., 2004). To limit the development of conditioned inhibition in second-order conditioning, fewer S2→S1 pairings should be employed. This is possible in conditioned taste aversion. Indeed, a single pairing between a gustatory S2 and a contextual S1 was sufficient to obtain sensory preconditioning and second-order conditioning, provided the US used to condition S1 was very salient (i.e., LiCl; Archer and Sjöden, 1982). Some instances of second-order fear conditioning consist of reinforced serial S2→S1 pairings following S1 training [i.e., S2→S1→US; Williams-Spooner et al., 2019; see also Mahmud et al. (2019)]. This design, like the standard non-reinforced design, results in robust learning about the second-order stimulus relative to an unpaired control (Leidl et al., 2018; Williams-Spooner et al., 2019). In reward learning, reinforced serial S2→S1 presentations lead to a higher level of responding during training compared to non-reinforced S2→S1 presentations (Holland, 1980). This effect, however, was likely due to the development of S2→US associations (Holland, 1980). To show this, Holland (1980) tested S2 under conditions that reveal the strength of second-order associations (i.e., under food satiation) and reported a lower level of responding to S2 when trained in the reinforced serial case. Holland (1980) further showed that surprising food presentations or omissions were more detrimental to second-order conditioning than when such events were expected.
- Stimulus Modality and Type: Various stimuli have been used in higher-order conditioning experiments, including color (Rashotte et al., 1977), shape (Rescorla, 1980a), odor (Holland, 1981), flavor (Holland, 1981, 1983), auditory cues such as tone (Rizley and Rescorla, 1972), white noise (Holland and Ross, 1983), clicker (Ward-Robinson and Hall, 1998), and visual cues such as key light (Rashotte et al., 1977), flashing light (Parkes and Westbrook, 2010; Wong et al., 2019), and context (Archer and Sjöden, 1982; Helmstetter and Fanselow, 1989; Iordanova et al., 2008). The types of USs used in higher-order designs are similar to those used in first-order conditioning studies, including footshocks, rewards such as food to a hungry rat, and lithium chloride (LiCl)-induced illness (Rizley and Rescorla, 1972; Holland and Rescorla, 1975a; Archer and Sjöden, 1982; Ward-Robinson and Hall, 1998). In first-order conditioning, an aversive US (e.g., a mild electric shock) conditions species-specific defensive behaviors (e.g., freezing; Blanchard and Blanchard, 1969; Bolles, 1970; Fanselow, 1980) or conditioned suppression (Rescorla and Furrow, 1977; Bouton and Bolles, 1980), whereas an appetitive US (e.g., sucrose pellets) supports conditioned approach (Holland, 1977). The US, however, is not the only determinant of conditioned responses. Auditory and visual cues can support cue-based responses including rearing, head jerk, perambulation, and general activity (Holland and Rescorla, 1975a,b; Holland, 1977, 1984). While auditory stimuli elicit startle and head jerk, visual stimuli elicit rearing (Holland, 1977). Startle and rearing are considered orienting responses (OR) and are seen to novel but not familiar non-reinforced cues and maintained or augmented to cues that have undergone conditioning. Head jerk is specific to conditioned auditory cues. ORs and CRs are differentially distributed across the duration of a conditioned stimulus, with ORs occurring mostly during the beginning of visual cues and food-cup CRs following afterward, while CRs and ORs elicited by auditory cues are more evenly distributed (Holland, 1977; Hatfield et al., 1996).
The associative links that govern sensory preconditioning and second-order conditioning differ depending on the procedural details. As different designs are often used to study the neural substrates of higher-order learning, it is imperative that one is aware that procedural differences can lead to differences in associative content (i.e., what is learned; see also Gewirtz and Davis, 2000; Parkes and Westbrook, 2011; Gostolupce et al., 2021). The first evidence to highlight the differences in learning between sensory preconditioning and second-order conditioning came from Rizley and Rescorla (1972). In a fear conditioning procedure with footshock as the US, the authors showed that reduction in responding to a stimulus (S1) that had been conditioned to an aversive US also reduced responding to a stimulus (S2) that had been previously paired with S1 in a sensory preconditioning design, but not in a second-order conditioning design. This indicated that the nature of the learned association differed between the two procedures.
The Role of Associative Learning in Mental Health
While associative learning is adaptive and generally beneficial, it can also play a significant role in the development of mental health problems. This typically happens when an individual forms maladaptive associations that lead to distressing or harmful behaviors.
- Anxiety Disorders: Anxiety can develop when an individual associates certain stimuli with fear or discomfort. For instance, if someone has a panic attack in a crowded place, they may start to associate crowds with fear, leading to avoidance behavior. Over time, this can lead to conditions such as agoraphobia, where individuals avoid places they associate with past anxiety attacks.
- Post-Traumatic Stress Disorder (PTSD): PTSD often involves strong associations between traumatic events and seemingly neutral stimuli. For example, a soldier might associate the sound of fireworks with battlefield explosions, which can trigger intense fear, flashbacks, and avoidance of situations where similar sounds might occur.
- Substance Abuse: Associative learning also plays a role in the development of addiction. Individuals often associate certain environments, people, or emotional states with drug use. These associations can lead to cravings and relapse when a person is exposed to these triggers, even after they have stopped using the substance.
- Phobias: Phobias are often rooted in associative learning, where an individual associates a specific object or situation with intense fear. This could develop from a direct negative experience (like being bitten by a dog) or through observation (seeing someone else react fearfully to dogs).
- Depression: While depression is complex and involves many factors, associative learning can contribute by reinforcing negative thought patterns and behaviors. For example, if an individual consistently experiences negative outcomes when attempting social interactions, they may begin to associate these situations with failure or rejection, reinforcing feelings of hopelessness.
Reshaping Maladaptive Associations Through Applied Psychology
Fortunately, several therapeutic approaches in applied psychology can help individuals reshape maladaptive associations, reducing their impact on mental health.
- Cognitive Behavioural Therapy (CBT): CBT is a widely used therapeutic approach that helps individuals identify and change maladaptive thought patterns and behaviors. By examining and challenging the associations that contribute to anxiety, depression, or other mental health issues, CBT can help individuals form healthier connections. For example, someone with social anxiety might learn to question the belief that others are judging them negatively in social situations. Through repeated practice, they can start to replace this association with a more positive and realistic understanding, reducing their anxiety over time.
- Exposure Therapy: This approach is particularly effective for anxiety disorders, PTSD, and phobias. It involves gradually exposing individuals to the feared stimulus in a controlled and safe environment, helping them break the association between the stimulus and their fear response. A person with a fear of flying, for example, might start by looking at pictures of airplanes, then visit an airport, and eventually take a short flight. Through repeated exposure without negative consequences, they can learn to associate flying with feelings of safety instead of fear.
- Systematic Desensitisation: This is a specific form of exposure therapy combined with relaxation techniques, particularly useful for phobias and other anxiety-related disorders. In systematic desensitization, a person learns relaxation techniques and gradually faces their fear hierarchy, starting from the least to the most anxiety-provoking situations. For instance, someone with a spider phobia might first learn deep breathing exercises, then look at pictures of spiders while practicing relaxation techniques, and eventually work up to being in the same room as a spider. This process can gradually weaken the fearful association and help the individual respond more calmly.
- Contingency Management for Addiction: This approach is often used in addiction treatment and uses positive reinforcement to encourage sobriety. By associating drug-free behavior with tangible rewards, individuals can form new positive associations that replace those linked to substance use. For example, in a treatment program, a person might receive a voucher for every negative drug test result. This positive reinforcement helps build a new association between abstinence and rewards, making it easier for the individual to resist cravings and avoid relapse.
Read also: Insightful Learning Explained
tags: #define #associative #learning
